Instance segmentation in dummy images

Abstract

The system is configured to segment an instance of the object in the dummy image. The dummy image may be generated from a 3D image coupled to the vehicle. A system may receive a dummy image, which may include a plurality of segments. The system may determine an object classification for the segment and determine an instance portion classification for the segment. The system may group a segment with another segment based on the object classification and the instance portion classification. The packet may correspond to an instance of the object in the image. The system may use the packet to navigate the vehicle.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. Patent Application No. 17 / 657,590, filed on March 31, 2022, entitled “INSTANCE SEGMENTATION IN APSEUDO-IMAGE,” which claims priority to U.S. Provisional Patent Application No. 63 / 269,601, filed on March 18, 2022, entitled “PROPOSAL-FREE LIDAR PANOPTIC SEGMENTATION WITH PILLAR-LEVEL AFFINITY,” each of which is incorporated herein by reference in its entirety. Background Art

[0003] Autonomous vehicle driving in complex environments poses significant technical challenges. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] Figure 1 is an example environment in which a vehicle including one or more components of an autonomous system may be implemented.

[0005] Figure 2 is a diagram of one or more systems of a vehicle including an autonomous system.

[0006] Figure 3 yes Figure 1 and Figure 2 A diagram of one or more devices and / or components of one or more systems.

[0007] Figure 4A is a diagram of some components of an autonomous system.

[0008] Figure 4B It is a diagram of the implementation of a neural network.

[0009] Figure 4C and Figure 4D is a diagram illustrating an example operation of a CNN.

[0010] Figure 5A is a block diagram illustrating an example of one or more components of a perception system.

[0011] Figure 5B is a block diagram illustrating an example of a pseudo image system.

[0012] Figure 5C is a block diagram illustrating an example of a point cloud and a column.

[0013] Figure 5D Example of a 2D ground plane illustrating a point cloud.

[0014] Fig. 6A and Figure 6B is an operational diagram illustrating example operations of perception system 402 to identify an instance of an agent in a pseudo image.

[0015] Figure 7 is a flow diagram illustrating an example of a routine implemented by one or more processors to segment an instance of an object. DETAILED DESCRIPTION

[0016] In the following description, for the purpose of explanation, many specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it will be apparent that the embodiments described in the present disclosure can be implemented without these specific details. In some instances, well-known configurations and devices are illustrated in block diagram form to avoid unnecessarily obscuring aspects of the present disclosure.

[0017] In the accompanying drawings, for ease of description, the specific arrangement or order of schematic elements (such as those representing systems, devices, modules, instruction blocks and / or data elements, etc.) is illustrated. However, those skilled in the art will understand that, unless explicitly described, the specific order or arrangement of schematic elements in the accompanying drawings is not intended to mean that a specific processing order or sequence, or separation of processing is required. In addition, unless explicitly described, the inclusion of schematic elements in the accompanying drawings is not intended to mean that such elements are required in all embodiments, nor is it intended to mean that the features represented by such elements cannot be included in some embodiments or cannot be combined with other elements in some embodiments.

[0018] In addition, in the accompanying drawings, connecting elements (such as solid or dotted lines or arrows, etc.) are used to illustrate the connection, relationship or association between or among two or more other schematic elements, and the absence of any such connecting elements is not intended to mean that there can be no connection, relationship or association. In other words, some connections, relationships or associations between elements are not illustrated in the accompanying drawings so as not to obscure the present disclosure. In addition, for ease of illustration, a single connecting element can be used to represent multiple connections, relationships or associations between elements. For example, if the connecting element represents the communication of a signal, data or instruction (e.g., "software instruction"), it will be understood by those skilled in the art that such an element can represent one or more than one signal path (e.g., bus) that may be needed to affect the communication.

[0019] Although the terms "first", "second" and / or "third", etc. are used to describe various elements, these elements should not be limited by these terms. The terms "first", "second" and / or "third" are only used to distinguish one element from another. For example, a first contact may be referred to as a second contact, and similarly, a second contact may be referred to as a first contact without departing from the scope of the described embodiments. Both the first contact and the second contact are contacts, but they are not the same contacts.

[0020] The terms used in the description of the various embodiments described herein are included only for the purpose of describing specific embodiments and are not intended to be limiting. As used in the description of the various embodiments described and in the appended claims, the singular forms "a", "an" and "the" are also intended to include plural forms and can be used interchangeably with "one or more than one" or "at least one", unless the context clearly indicates otherwise. It will also be understood that the term "and / or" as used herein refers to and includes any and all possible combinations of one or more than one of the associated listed items. It will also be understood that when the terms "include", "comprises", "have" and / or "have" are used in this specification, the stated features, integers, steps, operations, elements and / or components are specifically stated, but the presence or addition of one or more than one other features, integers, steps, operations, elements, components and / or groups thereof are not excluded.

[0021] As used herein, the terms "communication" and "communicating" refer to at least one of receiving, receiving, transmitting, transmitting and / or providing information (or information represented by, for example, data, signals, messages, instructions and / or commands, etc.). For a unit (e.g., a device, a system, a component of a device or system, and / or a combination thereof) to communicate with another unit, this means that the unit is able to directly or indirectly receive information from the other unit and / or send (e.g., transmit) information to the other unit. This can refer to a direct or indirect connection that is wired and / or wireless in nature. In addition, even if the transmitted information can be modified, processed, relayed and / or routed between the first unit and the second unit, the two units can communicate with each other. For example, even if the first unit passively receives information and does not actively transmit information to the second unit, the first unit can communicate with the second unit. As another example, if at least one intermediary unit (e.g., a third unit located between the first unit and the second unit) processes the information received from the first unit and transmits the processed information to the second unit, the first unit can communicate with the second unit. In some embodiments, a message may refer to a network packet (eg, a data packet, etc.) that includes data.

[0022] As used herein, the term "if" is optionally interpreted to mean "when," "at," "in response to being determined to be," and / or "in response to being detected," etc., depending on the context. Similarly, the phrases "if it is determined" or "if [the stated condition or event] is detected" are optionally interpreted to mean "when determining," "in response to being determined to be" or "when [the stated condition or event] is detected," and / or "in response to being detected," etc., depending on the context. In addition, as used herein, the terms "have," "have," or "possess," etc. are intended to be open-ended terms. Furthermore, unless expressly stated otherwise, the phrase "based on" is intended to mean "based at least in part on."

[0023] Reference will now be made in detail to embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the various embodiments described. However, it will be apparent to one of ordinary skill in the art that the various embodiments described may be implemented without these specific details. In other cases, well-known methods, processes, components, circuits, and networks have not yet been described in detail in order not to unnecessarily obscure aspects of the embodiments.

[0024] General Overview

[0025] In some aspects and / or embodiments, the systems, methods, and computer program products described herein include and / or implement a system configured to segment instances of an object. In some cases, the system can classify different segments of an image based on an object in the segment and / or a portion of an object in the segment. The classification of the segment can be used to group the segment with other segments to provide an object instance.

[0026] With the implementation of the systems, methods, and computer program products described herein, autonomous vehicles can be improved. For example, the vehicle can use fewer computing resources to identify object instances more quickly. Considering the computing resource constraints of autonomous vehicles, this can represent a significant improvement over previous systems. In addition, by using fewer computing resources to identify object instances more quickly, the system can enable the autonomous vehicle to more quickly perceive its surroundings and determine a path through these surroundings.

[0027] Reference now Figure 1, illustrates an example environment 100 in which vehicles including autonomous systems and vehicles not including autonomous systems operate. As illustrated, the environment 100 includes vehicles 102a-102n, objects 104a-104n, routes 106a-106n, areas 108, vehicle-to-infrastructure (V2I) devices 110, a network 112, a remote autonomous vehicle (AV) system 114, a fleet management system 116, and a V2I system 118. The vehicles 102a-102n, the vehicle-to-infrastructure (V2I) devices 110, the network 112, the autonomous vehicle (AV) system 114, the fleet management system 116, and the V2I system 118 are interconnected (e.g., establish connections for communication, etc.) via wired connections, wireless connections, or a combination of wired or wireless connections. In some embodiments, objects 104a-104n are interconnected with at least one of vehicles 102a-102n, vehicle-to-infrastructure (V2I) devices 110, networks 112, autonomous vehicle (AV) systems 114, fleet management systems 116, and V2I systems 118 via wired connections, wireless connections, or a combination of wired or wireless connections.

[0028] Vehicles 102a-102n (individually referred to as vehicles 102 and collectively referred to as vehicles 102) include at least one device configured to transport goods and / or people. In some embodiments, vehicles 102 are configured to communicate with V2I devices 110, remote AV systems 114, fleet management systems 116, and / or V2I systems 118 via network 112. In some embodiments, vehicles 102 include cars, buses, trucks, and / or trains, etc. In some embodiments, vehicles 102 are similar to vehicles 200 described herein (see Figure 2 ). In some embodiments, vehicles 200 in the set of vehicles 200 are associated with an autonomous queue manager. In some embodiments, vehicles 102 travel along respective routes 106a-106n (individually referred to as routes 106 and collectively referred to as routes 106) as described herein. In some embodiments, one or more vehicles 102 include an autonomous system (e.g., an autonomous system that is the same as or similar to autonomous system 202).

[0029] Objects 104a-104n (individually referred to as objects 104 and collectively referred to as objects 104) include, for example, at least one vehicle, at least one pedestrian, at least one cyclist, and / or at least one structure (e.g., a building, a sign, a fire hydrant, etc.), etc. Each object 104 is stationary (e.g., located at a fixed location and over a period of time) or moves (e.g., has a speed and is associated with at least one trajectory). In some embodiments, objects 104 are associated with corresponding locations in area 108.

[0030] Routes 106a-106n (individually referred to as routes 106 and collectively referred to as routes 106) are each associated with (e.g., specify a series of actions (also referred to as trajectories) connecting states along which the AV can navigate. Each route 106 begins at an initial state (e.g., a state corresponding to a first spatiotemporal location and / or speed, etc.) and ends at a final target state (e.g., a state corresponding to a second spatiotemporal location different from the first spatiotemporal location) or a target zone (e.g., a subspace of acceptable states (e.g., terminal states)). In some embodiments, the first state includes a location where one or more individuals will board the AV, and the second state or zone includes one or more locations where one or more individuals boarding the AV will disembark. In some embodiments, routes 106 include multiple acceptable state sequences (e.g., multiple spatiotemporal location sequences) that are associated with (e.g., define multiple trajectories). In an example, routes 106 include only high-level actions or imprecise state locations, such as a series of connecting roads indicating a change of direction at a roadway intersection, etc. Additionally or alternatively, the route 106 may include more precise actions or states, such as, for example, specific target lanes or precise locations within lane regions and target speeds at those locations, etc. In an example, the route 106 includes a plurality of precise state sequences along at least one high-level action with a limited look-ahead horizon to an intermediate target, wherein a combination of consecutive iterations of the limited horizon state sequences cumulatively correspond to a plurality of trajectories that collectively form a high-level route terminating at a final target state or region.

[0031] The area 108 includes a physical area (e.g., a geographic region) in which the vehicle 102 can navigate. In an example, the area 108 includes at least one state (e.g., a country, a province, a separate state of a plurality of states included in a country, etc.), at least a portion of a state, at least one city, at least a portion of a city, etc. In some embodiments, the area 108 includes at least one named thoroughfare (referred to herein as a "road"), such as a highway, an interstate highway, a parkway, a city street, etc. Additionally or alternatively, in some examples, the area 108 includes at least one unnamed road, such as a driveway, a section of a parking lot, a section of an open space and / or undeveloped area, a dirt road, etc. In some embodiments, the road includes at least one lane (e.g., a portion of the road that the vehicle 102 can traverse). In an example, the road includes at least one lane associated with (e.g., identified based on) at least one lane marking line.

[0032] The vehicle-to-infrastructure (V2I) device 110 (sometimes referred to as a vehicle-to-everything (V2X) device) includes at least one device configured to communicate with the vehicle 102 and / or the V2I system 118. In some embodiments, the V2I device 110 is configured to communicate with the vehicle 102, the remote AV system 114, the queue management system 116, and / or the V2I system 118 via the network 112. In some embodiments, the V2I device 110 includes a radio frequency identification (RFID) device, a sign, a camera (e.g., a two-dimensional (2D) and / or three-dimensional (3D) camera), lane markings, street lights, parking meters, etc. In some embodiments, the V2I device 110 is configured to communicate directly with the vehicle 102. Additionally or alternatively, in some embodiments, the V2I device 110 is configured to communicate with the vehicle 102, the remote AV system 114, and / or the queue management system 116 via the V2I system 118. In some embodiments, the V2I device 110 is configured to communicate with the V2I system 118 via the network 112 .

[0033] The network 112 includes one or more wired and / or wireless networks. In an example, the network 112 includes a cellular network (e.g., a long-term evolution (LTE) network, a third generation (3G) network, a fourth generation (4G) network, a fifth generation (5G) network, a code division multiple access (CDMA) network, etc.), a public land mobile network (PLMN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a telephone network (e.g., a public switched telephone network (PSTN)), a private network, an ad hoc network, an intranet, the Internet, a fiber-based network, a cloud computing network, etc., and / or a combination of some or all of these networks, etc.

[0034] The remote AV system 114 includes at least one device configured to communicate with the vehicle 102, the V2I device 110, the network 112, the fleet management system 116, and / or the V2I system 118 via the network 112. In an example, the remote AV system 114 includes a server, a server group, and / or other similar devices. In some embodiments, the remote AV system 114 is co-located with the fleet management system 116. In some embodiments, the remote AV system 114 participates in the installation of some or all of the components of the vehicle (including autonomous systems, autonomous vehicle computing, and / or software implemented by autonomous vehicle computing, etc.). In some embodiments, the remote AV system 114 maintains (e.g., updates and / or replaces) these components and / or software during the life of the vehicle.

[0035] The queue management system 116 includes at least one device configured to communicate with the vehicles 102, the V2I devices 110, the remote AV system 114, and / or the V2I system 118. In an example, the queue management system 116 includes a server, a server group, and / or other similar devices. In some embodiments, the queue management system 116 is associated with a ridesharing company (e.g., an organization for controlling the operation of multiple vehicles (e.g., vehicles including autonomous systems and / or vehicles not including autonomous systems)).

[0036] In some embodiments, the V2I system 118 includes at least one device configured to communicate with the vehicle 102, the V2I device 110, the remote AV system 114, and / or the fleet management system 116 via the network 112. In some examples, the V2I system 118 is configured to communicate with the V2I device 110 via a connection other than the network 112. In some embodiments, the V2I system 118 includes a server, a server group, and / or other similar devices. In some embodiments, the V2I system 118 is associated with a municipality or a private agency (e.g., a private agency for maintaining the V2I device 110, etc.).

[0037] supply Figure 1 The number and arrangement of elements illustrated are examples. Figure 1 There may be additional elements, fewer elements, different elements, and / or differently arranged elements than those illustrated. Additionally or alternatively, at least one element of environment 100 may be described as being Figure 1 Additionally or alternatively, at least one set of elements of environment 100 may perform one or more functions described as being performed by at least one different set of elements of environment 100.

[0038] Reference now Figure 2 , vehicle 200 includes autonomous system 202, powertrain control system 204, steering control system 206, and braking system 208. In some embodiments, vehicle 200 is compatible with vehicle 102 (see Figure 1 ) is the same or similar. In some embodiments, the vehicle 200 has autonomous capabilities (e.g., implementing at least one function, feature and / or device, etc., which enables the vehicle 200 to operate partially or completely without human intervention, including but not limited to fully autonomous vehicles (e.g., vehicles that abandon reliance on human intervention) and / or highly autonomous vehicles (e.g., vehicles that abandon reliance on human intervention in certain situations), etc.). For a detailed description of fully autonomous vehicles and highly autonomous vehicles, reference can be made to SAE International's standard J3016: Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle AutomatedDriving Systems, the entire contents of which are incorporated by reference. In some embodiments, the vehicle 200 is associated with an autonomous queue manager and / or a ride-sharing company.

[0039] Autonomous system 202 includes a sensor suite that includes one or more devices such as camera 202a, LiDAR sensor 202b, Radar sensor 202c, and microphone 202d. In some embodiments, autonomous system 202 may include more or fewer devices and / or different devices (e.g., ultrasonic sensors, inertial sensors, GPS receivers (discussed below), and / or odometer sensors for generating data associated with an indication of the distance that vehicle 200 has traveled, etc.). In some embodiments, autonomous system 202 uses one or more devices included in autonomous system 202 to generate data associated with environment 100 described herein. Data generated by one or more devices of autonomous system 202 can be used by one or more systems described herein to observe the environment (e.g., environment 100) in which vehicle 200 is located. In some embodiments, autonomous system 202 includes communication devices 202e, autonomous vehicle computing 202f, and a drive-by-wire (DBW) system 202h.

[0040] The camera 202a includes a communication device 202e, an autonomous vehicle computer 202f, and / or a safety controller 202g configured to communicate with the communication device 202e via a bus (e.g., Figure 3The camera 202a includes at least one device for communicating with the autonomous vehicle computing 202f (e.g., an image processing unit 202a, a bus ... Figure 1 In some embodiments, the autonomous vehicle computing 202f determines a depth to one or more objects in a field of view of at least two of the plurality of cameras based on image data from the at least two cameras. In some embodiments, the camera 202a is configured to capture images of objects within a distance relative to the camera 202a (e.g., up to 100 meters and / or up to 1 kilometer, etc.). Thus, the camera 202a includes features such as sensors and lenses that are optimized for sensing objects at one or more distances relative to the camera 202a.

[0041] In an embodiment, the camera 202a includes at least one camera configured to capture one or more images associated with one or more traffic lights, street signs, and / or other physical objects that provide visual navigation information. In some embodiments, the camera 202a generates traffic light data associated with the one or more images. In some examples, the camera 202a generates TLD data associated with one or more images including a format (e.g., RAW, JPEG, and / or PNG, etc.). In some embodiments, the camera 202a that generates TLD data differs from other systems incorporating cameras described herein in that the camera 202a may include one or more cameras with a wide field of view (e.g., a wide-angle lens, a fisheye lens, and / or a lens with a viewing angle of approximately 120 degrees or greater, etc.) to generate images related to as many physical objects as possible.

[0042] The light detection and ranging (LiDAR) sensor 202b includes a communication device 202e, an autonomous vehicle computing device 202f, and / or a safety controller 202g configured to communicate with the communication device 202e via a bus (e.g., Figure 3 The LiDAR sensor 202b includes at least one device that communicates with a bus (the same or similar bus as the bus 302 of the embodiment of the present invention). The LiDAR sensor 202b includes a system configured to emit light from a light emitter (e.g., a laser emitter). The light emitted by the LiDAR sensor 202b includes light outside the visible spectrum (e.g., infrared light, etc.). In some embodiments, during operation, the light emitted by the LiDAR sensor 202b encounters a physical object (e.g., a vehicle) and is reflected back to the LiDAR sensor 202b. In some embodiments, the light emitted by the LiDAR sensor 202b does not penetrate the physical object encountered by the light. The LiDAR sensor 202b also includes at least one light detector that detects the light emitted from the light emitter after encountering the physical object. In some embodiments, at least one data processing system associated with the LiDAR sensor 202b generates an image (e.g., a point cloud and / or a combined point cloud, etc.) representing objects included in the field of view of the LiDAR sensor 202b. In some examples, at least one data processing system associated with the LiDAR sensor 202b generates an image representing the boundaries of the physical object and / or the surface of the physical object (e.g., the topology of the surface), etc. In such examples, the image is used to determine the boundaries of the physical object in the field of view of the LiDAR sensor 202b.

[0043] The radio detection and ranging (Radar) sensor 202c includes a sensor configured to communicate with the communication device 202e, the autonomous vehicle computing device 202f, and / or the safety controller 202g via a bus (e.g., Figure 3At least one device for communicating with a bus (same or similar bus as bus 302 of the embodiment of the present invention). Radar sensor 202c includes a system configured to transmit (pulsed or continuous) radio waves. The radio waves transmitted by Radar sensor 202c include radio waves within a predetermined spectrum. In some embodiments, during operation, the radio waves transmitted by Radar sensor 202c encounter physical objects and are reflected back to Radar sensor 202c. In some embodiments, the radio waves transmitted by Radar sensor 202c are not reflected by some objects. In some embodiments, at least one data processing system associated with Radar sensor 202c generates a signal representing an object included in the field of view of Radar sensor 202c. For example, at least one data processing system associated with Radar sensor 202c generates an image representing the boundary of a physical object and / or the surface of a physical object (e.g., the topology of the surface), etc. In some examples, the image is used to determine the boundary of a physical object in the field of view of Radar sensor 202c.

[0044] The microphone 202d includes a microphone configured to communicate with the communication device 202e, the autonomous vehicle computing device 202f, and / or the safety controller 202g via a bus (e.g., Figure 3 At least one device for communicating with the vehicle 200 (the same or similar bus as bus 302 of FIG. 1 ). Microphone 202d includes one or more microphones (e.g., an array microphone and / or an external microphone, etc.) that capture an audio signal and generate data associated with (e.g., representing) the audio signal. In some examples, microphone 202d includes a transducer device and / or the like. In some embodiments, one or more systems described herein can receive the data generated by microphone 202d and determine the position (e.g., distance, etc.) of an object relative to vehicle 200 based on an audio signal associated with the data.

[0045] The communication device 202e includes at least one device configured to communicate with the camera 202a, the LiDAR sensor 202b, the Radar sensor 202c, the microphone 202d, the autonomous vehicle computing 202f, the safety controller 202g, and / or the drive-by-wire (DBW) system 202h. For example, the communication device 202e may include at least one device configured to communicate with the camera 202a, the LiDAR sensor 202b, the Radar sensor 202c, the microphone 202d, the autonomous vehicle computing 202f, the safety controller 202g, and / or the drive-by-wire (DBW) system 202h. Figure 3 The communication device 202e may be a device that is the same as or similar to the communication interface 314 of the vehicle. In some embodiments, the communication device 202e includes a vehicle-to-vehicle (V2V) communication device (eg, a device for enabling wireless communication of data between vehicles).

[0046] Autonomous vehicle computing 202f includes at least one device configured to communicate with camera 202a, LiDAR sensor 202b, Radar sensor 202c, microphone 202d, communication device 202e, safety controller 202g, and / or DBW system 202h. In some examples, autonomous vehicle computing 202f includes devices such as client devices, mobile devices (e.g., cellular phones and / or tablet computers, etc.), and / or servers (e.g., computing devices including one or more central processing units and / or graphics processing units, etc.). In some embodiments, autonomous vehicle computing 202f is the same or similar to autonomous vehicle computing 400 described herein. Additionally or alternatively, in some embodiments, autonomous vehicle computing 202f is configured to communicate with an autonomous vehicle system (e.g., with Figure 1 remote AV system 114 of the same or similar autonomous vehicle system), a fleet management system (e.g., Figure 1 The same or similar queue management system as the queue management system 116 of FIG. 1 ), V2I devices (e.g., Figure 1 V2I device 110 that is the same as or similar to V2I device 110) and / or V2I system (e.g., Figure 1 The V2I system 118 may communicate with the same or similar V2I system.

[0047] Safety controller 202g includes at least one device configured to communicate with camera 202a, LiDAR sensor 202b, Radar sensor 202c, microphone 202d, communication device 202e, autonomous vehicle computing 202f, and / or DBW system 202h. In some examples, safety controller 202g includes one or more controllers (electrical controllers and / or electromechanical controllers, etc.) configured to generate and / or transmit control signals to operate one or more devices of vehicle 200 (e.g., powertrain control system 204, steering control system 206, and / or braking system 208, etc.). In some embodiments, safety controller 202g is configured to generate control signals that take precedence over (e.g., override) control signals generated and / or transmitted by autonomous vehicle computing 202f.

[0048] The DBW system 202h includes at least one device configured to communicate with the communication device 202e and / or the autonomous vehicle computing 202f. In some examples, the DBW system 202h includes one or more controllers (e.g., electrical controllers and / or electromechanical controllers, etc.) configured to generate and / or transmit control signals to operate one or more devices of the vehicle 200 (e.g., powertrain control system 204, steering control system 206, and / or braking system 208, etc.). Additionally or alternatively, one or more controllers of the DBW system 202h are configured to generate and / or transmit control signals to operate at least one different device of the vehicle 200 (e.g., turn signals, headlights, door locks, and / or windshield wipers, etc.).

[0049] The powertrain control system 204 includes at least one device configured to communicate with the DBW system 202h. In some examples, the powertrain control system 204 includes at least one controller and / or actuator, etc. In some embodiments, the powertrain control system 204 receives a control signal from the DBW system 202h, and the powertrain control system 204 causes the vehicle 200 to start moving forward, stop moving forward, start moving backward, stop moving backward, accelerate in a certain direction, decelerate in a certain direction, turn left and / or turn right, etc. In an example, the powertrain control system 204 increases, maintains the same, or decreases the energy (e.g., fuel and / or electricity, etc.) provided to the motor of the vehicle, thereby rotating or not rotating at least one wheel of the vehicle 200.

[0050] The steering control system 206 includes at least one device configured to rotate one or more wheels of the vehicle 200. In some examples, the steering control system 206 includes at least one controller and / or actuator, etc. In some embodiments, the steering control system 206 rotates the two front wheels and / or the two rear wheels of the vehicle 200 to the left or right to turn the vehicle 200 left or right.

[0051] Braking system 208 includes at least one device configured to actuate one or more brakes to decelerate and / or hold vehicle 200 stationary. In some examples, braking system 208 includes at least one controller and / or actuator configured to cause one or more calipers associated with one or more wheels of vehicle 200 to close on respective rotors of vehicle 200. Additionally or alternatively, in some examples, braking system 208 includes an automatic emergency braking (AEB) system and / or a regenerative braking system, etc.

[0052] In some embodiments, vehicle 200 includes at least one platform sensor (not explicitly illustrated) for measuring or inferring a property of a state or condition of vehicle 200. In some examples, vehicle 200 includes platform sensors such as a global positioning system (GPS) receiver, an inertial measurement unit (IMU), wheel rate sensors, wheel brake pressure sensors, wheel torque sensors, engine torque sensors, and / or steering angle sensors.

[0053] Reference now Figure 3 , a schematic diagram of an exemplary device 300. As illustrated, device 300 includes a processor 304, a memory 306, a storage component 308, an input interface 310, an output interface 312, a communication interface 314, and a bus 302. In some embodiments, device 300 corresponds to: at least one device of vehicle 102 (e.g., at least one device of a system of vehicle 102); vehicle 200; and / or one or more devices of network 112 (e.g., one or more devices of a system of network 112). In some embodiments, one or more devices of vehicle 102 (e.g., one or more devices of a system of vehicle 102), vehicle 200, and / or one or more devices of network 112 (e.g., one or more devices of a system of network 112) include at least one device 300 and / or at least one component of device 300. As Figure 3 As shown, apparatus 300 includes a bus 302 , a processor 304 , a memory 306 , a storage component 308 , an input interface 310 , an output interface 312 , and a communication interface 314 .

[0054] The bus 302 includes components that permit communication between components of the device 300. In some cases, the processor 304 includes a processor (e.g., a central processing unit (CPU), a graphics processing unit (GPU), and / or an accelerated processing unit (APU), etc.), a microphone, a digital signal processor (DSP), and / or any processing component that can be programmed to perform at least one function (e.g., a field programmable gate array (FPGA) and / or an application specific integrated circuit (ASIC), etc.). The memory 306 includes a random access memory (RAM), a read-only memory (ROM), and / or another type of dynamic and / or static storage device (e.g., flash memory, magnetic memory, and / or optical memory, etc.) that stores data and / or instructions for use by the processor 304.

[0055] Storage component 308 stores data and / or software related to the operation and use of device 300. In some examples, storage component 308 includes a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optical disk, and / or a solid-state disk, etc.), a compact disk (CD), a digital versatile disk (DVD), a floppy disk, a cassette, a tape, a CD-ROM, a RAM, a PROM, an EPROM, a FLASH-EPROM, an NV-RAM, and / or another type of computer-readable medium, and a corresponding drive.

[0056] The input interface 310 includes components that permit the device 300 to receive information, such as via user input (e.g., a touch screen display, a keyboard, a keypad, a mouse, a button, a switch, a microphone, and / or a camera, etc.). Additionally or alternatively, in some embodiments, the input interface 310 includes a sensor for sensing information (e.g., a global positioning system (GPS) receiver, an accelerometer, a gyroscope, and / or an actuator, etc.). The output interface 312 includes components for providing output information from the device 300 (e.g., a display, a speaker, and / or one or more light emitting diodes (LEDs), etc.).

[0057] In some embodiments, communication interface 314 includes a transceiver-like component (e.g., a transceiver and / or a separate receiver and transmitter, etc.) that permits device 300 to communicate with other devices via a wired connection, a wireless connection, or a combination of a wired connection and a wireless connection. In some examples, communication interface 314 permits device 300 to receive information from another device and / or provide information to another device. In some examples, communication interface 314 includes an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a universal serial bus (USB) interface, interface and / or cellular network interface, etc.

[0058] In some embodiments, the device 300 performs one or more processes described herein. The device 300 performs these processes based on the processor 304 executing software instructions stored by a computer-readable medium such as a memory 306 and / or a storage component 308. Computer-readable media (e.g., non-transitory computer-readable media) are defined herein as non-transitory memory devices. Non-transitory memory devices include storage space located within a single physical storage device or storage space distributed across multiple physical storage devices.

[0059] In some embodiments, the software instructions are read into the memory 306 and / or storage component 308 from another computer-readable medium or from another device via the communication interface 314. The software instructions stored in the memory 306 and / or storage component 308, when executed, cause the processor 304 to perform one or more processes described herein. Additionally or alternatively, hardwired circuitry is used in place of or in combination with the software instructions to perform one or more processes described herein. Therefore, unless expressly stated otherwise, the embodiments described herein are not limited to any specific combination of hardware circuitry and software.

[0060] The memory 306 and / or the storage component 308 include a data storage unit or at least one data structure (e.g., a database, etc.). The device 300 can receive information from the data storage unit or at least one data structure in the memory 306 or the storage component 308, store information in the data storage unit or at least one data structure, communicate information to the data storage unit or at least one data structure, or search for information stored in the data storage unit or at least one data structure. In some examples, the information includes network data, input data, output data, or any combination thereof.

[0061] In some embodiments, the device 300 is configured to execute software instructions stored in the memory 306 and / or the memory of another device (e.g., another device that is the same as or similar to the device 300). As used herein, the term "module" refers to at least one instruction stored in the memory 306 and / or the memory of another device, which, when executed by the processor 304 and / or the processor of another device (e.g., another device that is the same as or similar to the device 300), causes the device 300 (e.g., at least one component of the device 300) to perform one or more processes described herein. In some embodiments, the module is implemented in software, firmware, and / or hardware, etc.

[0062] supply Figure 3 The number and arrangement of components illustrated are examples. Figure 3 The device 300 may include additional components, fewer components, different components, or differently arranged components than those illustrated. Additionally or alternatively, a set of components (e.g., one or more components) of the device 300 may perform one or more functions described as being performed by another component or set of components of the device 300.

[0063] Reference now Figure 4A, illustrates an example block diagram of an autonomous vehicle computing 400 (sometimes referred to as an "AV stack"). As illustrated, the autonomous vehicle computing 400 includes a perception system 402 (sometimes referred to as a perception module), a planning system 404 (sometimes referred to as a planning module), a positioning system 406 (sometimes referred to as a positioning module), a control system 408 (sometimes referred to as a control module), and a database 410. In some embodiments, the perception system 402, the planning system 404, the positioning system 406, the control system 408, and the database 410 are included in and / or implemented in an automatic navigation system of a vehicle (e.g., the autonomous vehicle computing 202f of the vehicle 200). Additionally or alternatively, in some embodiments, the perception system 402, the planning system 404, the positioning system 406, the control system 408, and the database 410 are included in one or more independent systems (e.g., one or more systems that are the same or similar to the autonomous vehicle computing 400, etc.). In some examples, the perception system 402, planning system 404, positioning system 406, control system 408, and database 410 are included in one or more independent systems located in the vehicle and / or at least one remote system as described herein. In some embodiments, any and / or all of the systems included in the autonomous vehicle computing 400 are implemented in software (e.g., software instructions stored in a memory), computer hardware (e.g., by a microprocessor, microcontroller, application specific integrated circuit (ASIC) and / or field programmable gate array (FPGA), etc.), or a combination of computer software and computer hardware. It will also be understood that in some embodiments, the autonomous vehicle computing 400 is configured to communicate with a remote system (e.g., an autonomous vehicle system that is the same or similar to the remote AV system 114, a fleet management system 116 that is the same or similar to the fleet management system 116, and / or a V2I system that is the same or similar to the V2I system 118, etc.).

[0064] In some embodiments, the perception system 402 receives data associated with at least one physical object in the environment (e.g., data used by the perception system 402 to detect at least one physical object) and classifies the at least one physical object. In some examples, the perception system 402 receives image data captured by at least one camera (e.g., camera 202a), the image being associated with (e.g., representing) one or more physical objects within the field of view of the at least one camera. In such examples, the perception system 402 classifies the at least one physical object based on one or more groups of physical objects (e.g., bicycles, vehicles, traffic signs, and / or pedestrians, etc.). In some embodiments, based on the classification of the physical object by the perception system 402, the perception system 402 transmits data associated with the classification of the physical object to the planning system 404.

[0065] In some embodiments, planning system 404 receives data associated with a destination and generates data associated with at least one route (e.g., route 106) along which a vehicle (e.g., vehicle 102) can travel toward the destination. In some embodiments, planning system 404 periodically or continuously receives data (e.g., data associated with the classification of physical objects described above) from perception system 402, and planning system 404 updates at least one trajectory or generates at least one different trajectory based on the data generated by perception system 402. In some embodiments, planning system 404 receives data associated with an updated position of a vehicle (e.g., vehicle 102) from positioning system 406, and planning system 404 updates at least one trajectory or generates at least one different trajectory based on the data generated by positioning system 406.

[0066] In some embodiments, the positioning system 406 receives data associated with (e.g., representing) a location of a vehicle (e.g., vehicle 102) in an area. In some examples, the positioning system 406 receives LiDAR data associated with at least one point cloud generated by at least one LiDAR sensor (e.g., LiDAR sensor 202b). In some examples, the positioning system 406 receives data associated with at least one point cloud from multiple LiDAR sensors, and the positioning system 406 generates a combined point cloud based on each point cloud. In these examples, the positioning system 406 compares the at least one point cloud or the combined point cloud with a two-dimensional (2D) and / or three-dimensional (3D) map of the area stored in the database 410. Then, based on the positioning system 406 comparing the at least one point cloud or the combined point cloud with the map, the positioning system 406 determines the position of the vehicle in the area. In some embodiments, the map includes a combined point cloud of the area generated before the navigation of the vehicle. In some embodiments, the map includes, but is not limited to, a high-precision map of roadway geometry, a map describing the road network connectivity, a map describing the physical properties of the roadway (such as traffic speed, traffic volume, the number of vehicle and bicycle traffic lanes, lane width, lane traffic direction or the type and location of lane markings, or a combination thereof, etc.), and a map describing the spatial location of road features (such as crosswalks, traffic signs, or various types of other driving lights, etc.). In some embodiments, the map is generated in real time based on data received by the perception system.

[0067] In another example, positioning system 406 receives global navigation satellite system (GNSS) data generated by a global positioning system (GPS) receiver. In some examples, positioning system 406 receives GNSS data associated with a location of a vehicle in an area, and positioning system 406 determines the latitude and longitude of the vehicle in the area. In such an example, positioning system 406 determines the position of the vehicle in the area based on the latitude and longitude of the vehicle. In some embodiments, positioning system 406 generates data associated with the position of the vehicle. In some examples, based on the location of the vehicle determined by positioning system 406, positioning system 406 generates data associated with the position of the vehicle. In such an example, the data associated with the position of the vehicle include data associated with one or more semantic properties corresponding to the position of the vehicle.

[0068] In some embodiments, the control system 408 receives data associated with at least one trajectory from the planning system 404, and the control system 408 controls the operation of the vehicle. In some examples, the control system 408 receives data associated with at least one trajectory from the planning system 404, and the control system 408 controls the operation of the vehicle by generating and transmitting control signals to operate a powertrain control system (e.g., DBW system 202h and / or powertrain control system 204, etc.), a steering control system (e.g., steering control system 206), and / or a braking system (e.g., braking system 208). In the example, in the case where the trajectory includes a left turn, the control system 408 transmits a control signal to cause the steering control system 206 to adjust the steering angle of the vehicle 200, thereby causing the vehicle 200 to turn left. Additionally or alternatively, the control system 408 generates and transmits control signals to cause other devices of the vehicle 200 (e.g., headlights, turn signals, door locks, and / or windshield wipers, etc.) to change state.

[0069] In some embodiments, perception system 402, planning system 404, positioning system 406, and / or control system 408 implement at least one machine learning model (e.g., at least one multi-layer perceptron (MLP), at least one convolutional neural network (CNN), at least one recurrent neural network (RNN), at least one autoencoder, and / or at least one transformer, etc.). In some examples, perception system 402, planning system 404, positioning system 406, and / or control system 408 implement at least one machine learning model alone or in combination with one or more of the above systems. In some examples, perception system 402, planning system 404, positioning system 406, and / or control system 408 implement at least one machine learning model as part of a pipeline (e.g., a pipeline for identifying one or more objects located in an environment, etc.). The following is about FIG. 4B to FIG. 4D Includes examples of implementations of machine learning models.

[0070] Database 410 stores data transmitted to, received from, and / or updated by perception system 402, planning system 404, positioning system 406, and / or control system 408. In some examples, database 410 includes a storage component (e.g., a storage component) for storing data and / or software related to operations and using at least one system of autonomous vehicle computing 400. Figure 3In some embodiments, database 410 stores data associated with a 2D and / or 3D map of at least one area. In some examples, database 410 stores data associated with a 2D and / or 3D map of a portion of a city, portions of multiple cities, multiple cities, counties, states, and / or countries (State) (e.g., a country), etc. In such an example, a vehicle (e.g., a vehicle that is the same or similar to vehicle 102 and / or vehicle 200) can be driven along one or more drivable areas (e.g., single-lane roads, multi-lane roads, highways, remote roads, and / or off-road roads, etc.) and cause at least one LiDAR sensor (e.g., a LiDAR sensor that is the same or similar to LiDAR sensor 202b) to generate data associated with an image representing an object included in the field of view of the at least one LiDAR sensor.

[0071] In some embodiments, database 410 can be implemented across multiple devices. In some examples, database 410 includes a vehicle (e.g., a vehicle that is the same or similar to vehicle 102 and / or vehicle 200), an autonomous vehicle system (e.g., an autonomous vehicle system that is the same or similar to remote AV system 114), a fleet management system (e.g., a vehicle that is the same or similar to remote AV system 114), and a fleet management system (e.g., a vehicle that is the same or similar to remote AV system 114). Figure 1 The same or similar queue management system as the queue management system 116 of FIG. 1 and / or the V2I system (e.g., Figure 1 The V2I system 118 is the same as or similar to the V2I system 118).

[0072] Reference now Figure 4B , a diagram illustrating an implementation of a machine learning model. More specifically, a diagram illustrating an implementation of a convolutional neural network (CNN) 420. For purposes of illustration, the following description of CNN 420 will be with respect to implementing CNN 420 by perception system 402. However, it will be understood that in some examples, CNN 420 (e.g., one or more components of CNN 420) is implemented by other systems (such as planning system 404, positioning system 406, and / or control system 408, etc.) other than or in addition to perception system 402. Although CNN 420 includes certain features as described herein, these features are provided for purposes of illustration and are not intended to limit the present disclosure.

[0073] CNN 420 includes a plurality of convolutional layers including a first convolutional layer 422, a second convolutional layer 424, and a convolutional layer 426. In some embodiments, CNN 420 includes a subsampling layer 428 (sometimes referred to as a pooling layer). In some embodiments, subsampling layer 428 and / or other subsampling layers have a dimension that is smaller than the dimension of the upstream system (i.e., the number of nodes). With the subsampling layer 428 having a dimension that is smaller than the dimension of the upstream layer, CNN 420 merges the amount of data associated with the initial input and / or output of the upstream layer, thereby reducing the amount of computation required for CNN 420 to perform downstream convolution operations. Additionally or alternatively, with the subsampling layer 428 being associated with (e.g., configured to perform) at least one subsampling function (as described below with respect to Figure 4C and Figure 4D As described above, CNN 420 incorporates the amount of data associated with the initial input.

[0074] The perception system 402 performs the convolution operation based on the perception system 402 providing respective inputs and / or outputs associated with each of the first convolution layer 422, the second convolution layer 424, and the convolution layer 426 to generate respective outputs. In some examples, the perception system 402 implements the CNN 420 based on the perception system 402 providing data as input to the first convolution layer 422, the second convolution layer 424, and the convolution layer 426. In such examples, the perception system 402 provides data as input to the first convolution layer 422, the second convolution layer 424, and the convolution layer 426 based on the perception system 402 receiving data from one or more different systems (e.g., one or more systems of a vehicle that is the same or similar to the vehicle 102, a remote AV system that is the same or similar to the remote AV system 114, a queue management system that is the same or similar to the queue management system 116, and / or a V2I system that is the same or similar to the V2I system 118, etc.). The following is about Figure 4C Includes a detailed description of the convolution operation.

[0075] In some embodiments, the perception system 402 provides data associated with the input (referred to as the initial input) to the first convolutional layer 422, and the perception system 402 generates data associated with the output using the first convolutional layer 422. In some embodiments, the perception system 402 provides the output generated by the convolutional layer as input to a different convolutional layer. For example, the perception system 402 provides the output of the first convolutional layer 422 as input to the subsampling layer 428, the second convolutional layer 424, and / or the convolutional layer 426. In such an example, the first convolutional layer 422 is referred to as an upstream layer, and the subsampling layer 428, the second convolutional layer 424, and / or the convolutional layer 426 are referred to as downstream layers. Similarly, in some embodiments, the perception system 402 provides the output of the subsampling layer 428 to the second convolutional layer 424 and / or the convolutional layer 426, and in this example, the subsampling layer 428 will be referred to as the upstream layer, and the second convolutional layer 424 and / or the convolutional layer 426 will be referred to as the downstream layer.

[0076] In some embodiments, before the perception system 402 provides the input to the CNN 420, the perception system 402 processes the data associated with the input provided to the CNN 420. For example, the perception system 402 processes the data associated with the input provided to the CNN 420 based on the perception system 402 normalizing the sensor data (e.g., image data, LiDAR data, and / or Radar data, etc.).

[0077] In some embodiments, CNN 420 generates an output based on perception system 402 performing convolution operations associated with each convolution layer. In some examples, CNN 420 generates an output based on perception system 402 performing convolution operations associated with each convolution layer and the initial input. In some embodiments, perception system 402 generates an output and provides the output to fully connected layer 430. In some examples, perception system 402 provides the output of convolution layer 426 to fully connected layer 430, wherein fully connected layer 430 includes data associated with multiple feature values ​​referred to as F1, F2, ..., FN. In this example, the output of convolution layer 426 includes data associated with multiple output feature values ​​representing predictions.

[0078] In some embodiments, perception system 402 identifies a prediction from the plurality of predictions based on perception system 402 identifying a feature value associated with a highest likelihood of being a correct prediction from the plurality of predictions. For example, where fully connected layer 430 includes feature values ​​F1, F2, ..., FN and F1 is the largest feature value, perception system 402 identifies the prediction associated with F1 as the correct prediction from the plurality of predictions. In some embodiments, perception system 402 trains CNN 420 to generate the predictions. In some examples, perception system 402 trains CNN 420 to generate the predictions based on perception system 402 providing training data associated with the predictions to CNN 420.

[0079] Reference now Figure 4C and Figure 4D , a diagram illustrating an example operation of CNN 440 utilizing perception system 402. In some embodiments, CNN 440 (e.g., one or more components of CNN 440) is coupled to CNN 420 (e.g., one or more components of CNN 420) (see Figure 4B ) are the same or similar.

[0080] At step 450, the perception system 402 provides data associated with the image as input to the CNN 440 (step 450). For example, as illustrated, the perception system 402 provides data associated with the image to the CNN 440, where the image is a grayscale image represented as values ​​stored in a two-dimensional (2D) array. In some embodiments, the data associated with the image may include data associated with a color image represented as values ​​stored in a three-dimensional (3D) array. Additionally or alternatively, the data associated with the image may include data associated with an infrared image and / or a Radar image, etc.

[0081] At step 455, CNN 440 performs a first convolution function. For example, CNN 440 performs a first convolution function based on CNN 440 providing a value representing an image as an input to one or more neurons (not explicitly illustrated) included in first convolution layer 442. In this example, the value representing the image may correspond to a value of a region (sometimes referred to as a receptive field) representing the image. In some embodiments, each neuron is associated with a filter (not explicitly illustrated). The filter (sometimes referred to as a kernel) may be represented as an array of values ​​corresponding in size to the value provided as input to the neuron. In one example, the filter may be configured to identify edges (e.g., horizontal lines, vertical lines, and / or straight lines, etc.). In successive convolution layers, the filters associated with the neurons may be configured to continuously identify more complex patterns (e.g., arcs and / or objects, etc.).

[0082] In some embodiments, CNN 440 performs a first convolution function based on CNN 440 multiplying the values ​​of each neuron provided as input to one or more neurons included in the first convolution layer 442 by the values ​​of the filters corresponding to each neuron in the same or more neurons. For example, CNN 440 may multiply the values ​​of each neuron provided as input to one or more neurons included in the first convolution layer 442 by the values ​​of the filters corresponding to each neuron in the one or more neurons to generate a single value or an array of values ​​as output. In some embodiments, the collective output of the neurons of the first convolution layer 442 is referred to as a convolution output. In some embodiments, when each neuron has the same filter, the convolution output is referred to as a feature map.

[0083] In some embodiments, CNN 440 provides the output of each neuron of the first convolutional layer 442 to the neurons of the downstream layer. For clarity, the upstream layer may be a layer that transmits data to a different layer (referred to as the downstream layer). For example, CNN 440 may provide the output of each neuron of the first convolutional layer 442 to the corresponding neurons of the subsampling layer. In the example, CNN 440 provides the output of each neuron of the first convolutional layer 442 to the corresponding neurons of the first subsampling layer 444. In some embodiments, CNN 440 adds a bias value to the aggregate set of all values ​​provided to each neuron of the downstream layer. For example, CNN 440 adds a bias value to the aggregate set of all values ​​provided to each neuron of the first subsampling layer 444. In such an example, CNN 440 determines the final value to be provided to each neuron of the first subsampling layer 444 based on the aggregate set of all values ​​provided to each neuron and the activation function associated with each neuron of the first subsampling layer 444.

[0084] At step 460, CNN 440 performs a first subsampling function. For example, based on CNN 440 providing the values ​​output by first convolutional layer 442 to the corresponding neurons of first subsampling layer 444, CNN 440 may perform the first subsampling function. In some embodiments, CNN 440 performs the first subsampling function based on an aggregation function. In an example, CNN 440 performs the first subsampling function based on CNN 440 determining the maximum input (referred to as a maximum pooling function) among the values ​​provided to a given neuron. In another example, CNN 440 performs the first subsampling function based on CNN 440 determining the average input (referred to as an average pooling function) among the values ​​provided to a given neuron. In some embodiments, based on CNN 440 providing values ​​to the respective neurons of first subsampling layer 444, CNN 440 generates an output, which is sometimes referred to as a subsampled convolution output.

[0085] At step 465, CNN 440 performs a second convolution function. In some embodiments, CNN 440 performs the second convolution function in a manner similar to how CNN 440 performs the first convolution function described above. In some embodiments, CNN 440 performs the second convolution function based on CNN 440 providing the value output by first subsampling layer 444 as input to one or more neurons (not explicitly illustrated) included in second convolution layer 446. In some embodiments, as described above, each neuron of second convolution layer 446 is associated with a filter. As described above, the filter (one or more) associated with second convolution layer 446 can be configured to recognize more complex patterns than the filter associated with first convolution layer 442.

[0086] In some embodiments, the CNN 440 performs a second convolution function based on the CNN 440 multiplying the value of each neuron provided as input to the one or more neurons included in the second convolution layer 446 by the value of the filter corresponding to each neuron of the one or more neurons. For example, the CNN 440 may multiply the value of each neuron provided as input to the one or more neurons included in the second convolution layer 446 by the value of the filter corresponding to each neuron of the one or more neurons to generate a single value or a value array as an output.

[0087] In some embodiments, the CNN 440 provides the output of each neuron of the second convolutional layer 446 to the neurons of the downstream layer. For example, the CNN 440 may provide the output of each neuron of the first convolutional layer 442 to the corresponding neurons of the subsampling layer. In an example, the CNN 440 provides the output of each neuron of the first convolutional layer 442 to the corresponding neurons of the second subsampling layer 448. In some embodiments, the CNN 440 adds a bias value to the aggregate set of all values ​​provided to each neuron of the downstream layer. For example, the CNN 440 adds a bias value to the aggregate set of all values ​​provided to each neuron of the second subsampling layer 448. In such an example, the CNN 440 determines the final value provided to each neuron of the second subsampling layer 448 based on the aggregate set of all values ​​provided to each neuron and the activation function associated with each neuron of the second subsampling layer 448.

[0088] At step 470, CNN 440 performs a second subsampling function. For example, based on CNN 440 providing the values ​​output by second convolutional layer 446 to corresponding neurons of second subsampling layer 448, CNN 440 may perform a second subsampling function. In some embodiments, based on CNN 440 using an aggregation function, CNN 440 performs a second subsampling function. In an example, as described above, based on CNN 440 determining the maximum input or average input among the values ​​provided to a given neuron, CNN 440 performs a first subsampling function. In some embodiments, based on CNN 440 providing values ​​to respective neurons of second subsampling layer 448, CNN 440 generates an output.

[0089] At step 475, CNN 440 provides outputs 480 of each neuron of second subsampling layer 448 to fully connected layer 449. For example, CNN 440 provides outputs of each neuron of second subsampling layer 448 to fully connected layer 449 so that fully connected layer 449 generates outputs. In some embodiments, fully connected layer 449 is configured to generate outputs associated with predictions (sometimes referred to as classifications). The predictions may include indications of objects included in the image provided as input to CNN 440, including objects and / or sets of objects, etc. In some embodiments, perception system 402 performs one or more operations and / or provides data associated with the predictions to various systems described herein.

[0090] General Overview

[0091] Autonomous vehicle driving in complex environments (e.g., urban environments) presents significant technical challenges. To navigate in these environments, vehicles use sensors such as LiDAR, Radar, cameras, or ultrasonic sensors to detect various types of objects such as vehicles, pedestrians, and bicycles in real time.

[0092] One approach for object detection on image input is deep learning. However, the sparsity of sensor data (e.g., LiDAR point clouds) makes existing image-based deep learning techniques computationally inefficient. The disclosed embodiments include systems and techniques for efficiently and quickly detecting objects based on 3D sensor data.

[0093] In some cases, using 3D sensor data to identify an object may include transforming a 3D image into a 2D pseudo image and performing semantic segmentation and instance portion segmentation on the 2D pseudo image. Semantic segmentation may include labeling portions of an image (e.g., pixels, data points, or segments of a pseudo image) and classification of different objects in the image. These objects may include: objects that move or act in some way, such as, but not limited to, bicycles, buses, cars, construction vehicles, motorcycles, pedestrians, trailers, trucks, strollers, or other vehicles, etc. (also referred to herein as "agents" or "agent objects"); and stationary, non-acting, or background objects (also referred to herein as "non-agents" or "non-agent objects"), such as, but not limited to, obstacles, traffic cones, drivable surfaces, other flat surfaces, sidewalks, terrain, man-made structures, vegetation, etc. Instance portion segmentation may include classification of portions of different segments of an image (non-limiting example: a column of a dot-column image) or instances of an object. In some cases, instance portion segmentation may classify segments of an image that include an agent object. The combination of semantic segmentation and instance part segmentation can be used to identify instances of different types of objects in an image (e.g., vehicle 1, vehicle 2, vehicle 3, pedestrian 1, pedestrian 2, bicycle 1, etc.) and can enable the vehicle to plan a route through the scene.

[0094] Classifying objects (e.g., semantic segmentation) as well as classifying parts of instances (e.g., instance part segmentation) can present significant technical challenges. For example, there can be conflicts between object classification and instances, inefficiencies can arise due to the two tasks sharing redundant information, and overlapping identifiers (e.g., boxes) can lead to over-identification of objects. Furthermore, identifying instances of intelligent objects using 3D sensor data can be complex and use significant computational resources, which can take more time than a vehicle can implement to sense its environment and determine a route through it.

[0095] The disclosed embodiments include systems and techniques for efficiently detecting instances of objects (and / or agents) from 3D sensor data. In some cases, the system can generate a pseudo image from the 3D sensor data and simultaneously classify parts of instances of agent objects (instance part segmentation) and classify objects (e.g., semantic segmentation) based on the pseudo image.

[0096] Using instance part classification and object classification, the system can traverse the pseudo image and group different segments of the pseudo image with different agent instances. For example, the system can determine whether various segments of the pseudo image correspond to existing agent instances or to new (or unrecognized) agent instances. In some cases, when the system traverses the pseudo image and encounters a segment of the pseudo image that is classified as an agent object, the system can determine whether the segment shares an instance of the agent with other segments of the pseudo image or whether the segment is the beginning of an instance of the agent. For example, in the case where the image is a dot column image, the system can scan the columns of the dot column image from left to right and determine whether the column being scanned is the beginning of an instance or a part of an instance with a column on the left.

[0097] By classifying instance parts and performing semantic segmentation on the pseudo-image, the system can reduce the computational complexity of clustering object instances. For example, in some cases, the system can cluster instances without performing complex distribution calculations on data points in the pseudo-image. In this way, the system can reduce the load and computational requirements of the vehicle's processing system, thereby improving the rate and accuracy of object detection in the vehicle scene.

[0098] Figure 5A is a block diagram illustrating an example of one or more components of perception system 402. In the illustrated example, perception system 402 includes pseudo image system 504, instance segmentation system 505, and one or more prediction systems 514. Further, in the illustrated example, instance segmentation system 505 includes image feature extraction network 506, semantic segmentation system 508, instance part segmentation system 510, and instance identifier system 512. In some cases, any one or any combination of pseudo image system 504, image feature extraction network 506, semantic segmentation system 508, instance identifier system 510, and one or more prediction systems 514 may be implemented using one or more neural networks and / or one or more processors or computer hardware (e.g., by a microprocessor, microcontroller, application-specific integrated circuit [ASIC, Application-Specific Integrated Circuit] and / or field programmable gate array (FPGA, Field Programmable Gate Array), etc.).

[0099] The pseudo image system 504 may receive the 3D image 502 (e.g., LiDAR, Radar, etc.) and generate a pseudo image. In some cases, the pseudo image may be a pseudo 2D image. In some cases, the pseudo image may be a Bird's Eye View (BEV) image of a scene of an autonomous vehicle. In some cases, the pseudo image system 504 may be implemented using one or more neural networks trained to receive 3D image data and output a pseudo image.

[0100] The instance segmentation system 505 can receive a pseudo image, extract features from the pseudo image, segment an object in the pseudo image, segment an instance part in the pseudo image, and identify an instance of an object in the pseudo image. In the illustrated example, the instance segmentation system is implemented using an image feature extraction network 506, a semantic segmentation system 508, an instance part segmentation system 510, and an instance identifier system 512, however, it will be understood that fewer or more components may be included in the instance segmentation system 505, and / or one or more of these components may be combined.

[0101] Prediction system(s) 514 may use output from instance identifier system 512 to determine one or more predictions for instances identified by instance identifier system 512. In some cases, prediction system(s) 514 may be implemented using one or more CNNs that are the same or similar to CNN 420 and / or CNN 440, and may be configured to receive instance identifiers and pseudo image data, and output feature data associated with one or more features extracted from the pseudo image, and / or output an annotated image that includes feature data associated with one or more features extracted from the received data and image data. In some cases, prediction system(s) 514 may be configured to determine the width, height, and length of an object in an image, a bounding box of an object, object movement, and / or object orientation. In some cases, prediction system(s) 514 are configured to determine features similar to LiDAR neural networks and object trajectory predictions.

[0102] Pseudo-image system

[0103] refer to Figure 5B, illustrating example components of a pseudo image system 504. In the illustrated example, the pseudo image system 504 includes a column creation component 556, an encoder 560, and a 2D image creation component 564. The column creation component 556 is configured to receive a set of measurements (e.g., a point cloud 553) from a sensor 552 of a vehicle as input. In some embodiments, the sensor 552 is a LiDAR (e.g., a LiDAR as described above). In some embodiments, the sensor 552 is a Radar (e.g., a Radar as described above). In an embodiment, the column creation component 556 is configured to receive a merged point cloud generated by combining point clouds from multiple LiDARs.

[0104] Point cloud 553 includes a plurality of data points representing a plurality of objects in 3D space around a vehicle. For example, the plurality of data points represent a plurality of objects, including one or more of a vehicle (e.g., a car, a bicycle, or a truck), a pedestrian, an animal, a static object (e.g., vegetation, a building, etc.), or an infrastructure (e.g., a traffic light). Each data point in the plurality of data points is a set of 3D space coordinates (e.g., (x, y, z) coordinates). Figure 5C An example point cloud with multiple data points is illustrated in .

[0105] The column creation component 556 is configured to divide the 3D space into a plurality of columns (also referred to herein as segments). Each of the plurality of columns is a slice of the 3D space, and each column extends from a corresponding portion of a 2D ground plane (e.g., an xy plane) of the 3D space. In an embodiment, the 3D space coordinates are defined relative to the LiDAR coordinate frame. The xy plane is parallel to the ground, and z is perpendicular to the ground. In an embodiment, the column extends infinitely upward and downward (z direction) corresponding to the area below the ground in the environment and toward the sky. For example, in the case of using a ground-penetrating Radar or other sensors that scan features under the surface, the column extends downward to below the surface. Similarly, if the sensor includes a LiDAR with a large field of view or other sensors that scan a large area above the ground. In an embodiment, the column has a fixed minimum height and a fixed maximum height corresponding to an observation environment including the ground and the highest object of interest.

[0106] In some embodiments, the column creation component 556 divides the 2D ground plane into a 2D grid including grid cells of equal size (e.g., square grid cells with sides of equal length), and therefore the columns extending vertically (e.g., in the z direction) in 3D space from these 2D grid cells have the same volume. In an embodiment, the size of the grid cells is variable and can be determined based on the computational requirements. A coarser grid will be less accurate and require less computational resources. Similarly, a finer grid will result in increased accuracy at the expense of increased computational resources. For example, consider a scene divided into a 2D grid of 50m×50m. If each column of the grid is 50cm×50cm, the 2D grid will have 10,000 columns or segments. If each column is 20cm×20cm, the 2D grid will have 62,500 columns or segments, which can result in increased accuracy and use more computational resources compared to a grid of 10,000 columns. In some cases, the grid cells have sides of unequal lengths.

[0107] In some cases, the pillar creation component 556 divides the 2D ground plane into 2D grids with different grid cell sizes. For example, the component 556 uses additional input 555 from other sensors 554 (e.g., cameras) to detect the density of objects in the point cloud 553. The component 556 can generate a 2D grid with different grid cell sizes so that there are more pillars in areas with high object density, fewer pillars in areas with low object density, and no pillars in areas without objects.

[0108] In some embodiments, the pillar creation component 556 divides the 2D ground plane into a 2D polar coordinate grid, as described in more detail in U.S. Provisional Application No. 63 / 194,694 filed on May 28, 2021, U.S. Provisional Application No. 63 / 191,887 filed on May 21, 2021, and U.S. Application No. Pending, Attorney Docket No. 46154-0381001 / I2021190, entitled "STREAMING OBJECTDETECTION AND SEGMENTATION WITH POLAR PILLARS" filed on March 31, 2022, each of which is incorporated herein by reference in its entirety for all purposes. In some cases, the 2D polar coordinate grid can have different grid cell sizes. For example, component 556 uses additional input 555 from other sensors 554 (e.g., cameras, LiDAR, Radar, or ultrasonic sensors) to detect the density of objects in the point cloud 553. Component 556 can generate a 2D polar coordinate grid in a polar coordinate system such that there are columns having a shape corresponding to the LiDAR coordinate frame (e.g., the natural shape of the point cloud data). In other words, most of the cells in the 2D polar coordinate grid contain data and are not empty due to the lack of correspondence between the shape of the point cloud data and the shape of the grid. In some embodiments, the 2D polar coordinate grid includes more columns in areas of high object density, fewer columns in areas of low object density, and no columns in areas without objects. In some embodiments, the 2D polar coordinate grid includes a plurality of substantially wedge-shaped cells. In the example, the point cloud 553 is divided into a plurality of (n) sectors according to the azimuth. For example, for a 2D polar coordinate grid divided into 32 segments, each sector corresponds to an azimuth of 360 / 32°.

[0109] Then, the column creation component 556 assigns each of the plurality of data points to a column in the plurality of columns. For example, each of the plurality of data points is assigned to a corresponding column based on the 2D coordinates of the data point. That is, if the 2D coordinates of the data point are within a particular portion of the 2D ground plane from which a particular column extends, then the data point is assigned to the particular column.

[0110] After the data points are assigned to the bins, the bin creation component 556 determines whether a first count of multiple non-empty bins (bins with at least one data point) exceeds a threshold value P.

[0111] If the first count of the plurality of non-empty columns exceeds P, the column creation component 556 selects P non-empty columns from the plurality of non-empty columns. For example, the column creation component 556 randomly subsamples P non-empty columns from the plurality of non-empty columns.

[0112] If the first count of non-empty bars is less than the first threshold value P, the bar creation component 556 generates a second subset of empty bars such that the sum of the first count and the second count of the second subset of bars is equal to P. The bar creation component 556 then selects P non-empty bars from the plurality of non-empty bars and the second subset of bars.

[0113] For each non-empty column in the P non-empty columns, the column creation component 556 is configured to maintain a threshold number of data points in each non-empty column. To this end, the column creation component 556 first determines whether the third count of data points assigned to the non-empty column exceeds the second threshold N for each non-empty column in the P non-empty columns.

[0114] For each non-empty column in the P non-empty columns, if the third count of data points assigned to the non-empty column exceeds a second threshold N, the column creation component 556 selects N data points to be maintained in the non-empty column. For example, the column creation component 556 randomly subsamples N data points from the data points assigned to the non-empty column. In an embodiment, N is typically determined to be high enough so that there is a near statistical certainty (>99%) that one or more points from each object are captured in the point column. In an embodiment, different algorithms are used to further reduce the risk of missing objects during the sampling of data points. If the third count of data points in the non-empty column is less than N, the column creation component 556 assigns multiple zero-coordinate data points to the non-empty column so that the sum of the fourth count of the multiple zero coordinates and the third count is equal to N.

[0115] In some embodiments, the first threshold value P and the second threshold value N are predetermined values. In an embodiment, P and N are predetermined based on the distribution of data points, so that a small portion of the data points are removed. In some embodiments, the first threshold value P and the second threshold value N are adaptive values. In particular, based on the density of objects in the 3D space, the column creation component 556 can adjust P and / or N so that there are more columns and / or more data points are allowed in each column in areas with high object density, there are fewer columns and / or fewer data points are allowed in each column in areas with low object density, and there are no columns in areas without objects.

[0116] For each non-empty column in the plurality of non-empty columns, the column creation component 556 generates a plurality of modified data points based on the plurality of data points in each non-empty column. In particular, for each non-empty column, the column creation component 556 generates a corresponding modified data point for each data point in the non-empty column based on the relative distance between the data point and the center of the non-empty column. The center of the column is selected so that the coordinate system of the modified data point is aligned with the neural network used in subsequent processing. In an embodiment, the column creation component 556 generates a corresponding modified data point for each data point in the non-empty column based on the relative distance between the data point and the center of the non-empty column, and further based on the cylindrical Euclidean distance from the sensor 552 to the data point. Then, the column creation component 556 transforms the plurality of data points in each non-empty column into a plurality of modified data points generated for the non-empty column. In an embodiment, for each non-empty column, the column creation component 556 generates a corresponding modified data point for each data point in the non-empty column based on the relative distance between the data point and the center of gravity (average position) of the points in the column.

[0117] For example, in an embodiment, each data point in the plurality of data points is represented by a 3D spatial coordinate (x, y, z), a reflectivity (r), and a timestamp (t). The column creation component 556 transforms each data point (x, y, z, r, t) in the non-empty column into a corresponding modified data point (x offset ,y offset ,z,r,t,d), where x offset (x 偏移 ) and y offset (y 偏移 ) is measured based on the relative distance between the data point and the center of the non-empty column, z is the height of the data point in the non-empty column, t is the timestamp, and d is a distance metric such as the cylindrical Euclidean distance from the sensor 552 to the data point. Other distance metrics are also possible. Each modified data point has D dimensions, where D is equal to the number of dimensions of the modified data point. In this example embodiment, each modified data point has D=6 dimensions: x offset ,y offset , z, r, t and d.

[0118] In an embodiment, the column creation component 556 assigns a column index to each non-empty column in the P non-empty columns and assigns a data point index to each modified data point in the P non-empty columns. The column creation component 556 generates a P-dimensional column index vector 559 that maps the column index of each column to a corresponding position (e.g., a corresponding grid cell) in the original 2D grid from which the column extends vertically. In some embodiments, the 2D grid is a 2D polar coordinate grid.

[0119] In an embodiment, the column creation component 556 generates a 3D stacked column tensor 557 for all non-empty columns and all modified data points, which is a (D, P, N)-dimensional tensor with modified data point coordinates, column index coordinates, and data point index coordinates. For each modified data point in the P non-empty columns, the 3D stacked column tensor 557 maps the column index of the column including the modified data point and the data point index of the modified data point to the modified data point.

[0120] The column creation component 556 then generates a column output 558 comprising a 3D stacked column tensor 557 and a P-dimensional column index vector 559.

[0121] In an embodiment, encoder 560 is a neural network configured to receive column output 558 and process 3D stacked column tensor 557 to generate learned feature output 562 that characterizes features of point cloud 553.

[0122] Specifically, to generate the learned feature output 562, the encoder 560 uses the 3D stacked column tensor 557 to initialize the current 3D feature tensor, and iteratively performs the following steps K times, where K is a predetermined number:

[0123] 1. Apply a 1×1 convolution across the modified data point index coordinates and column index coordinates of the current 3D feature tensor (i.e., across the (N, P) canvas) to generate a vector with size (C K ,N,P)’s first tensor T.

[0124] 2. Apply the element-wise maximum operator across the modified data point index coordinates of the first tensor T to generate a vector with size (C K ,P). In particular, the maximum matrix M is calculated across all modified data points in each column such that:

[0125]

[0126] 3. Determine whether the current iteration is the Kth iteration.

[0127] a) If the current iteration is the Kth iteration, the current maximum matrix M is output as the learned feature output 562 of the encoder 560. The output 562 is (C K ,P) tensor.

[0128] b) If the current iteration is not the Kth iteration, a second (C t) is generated by repeating the maximum matrix M N times along the second dimension of the first tensor T (ie, the modified data point index coordinates). K ,N,P) tensor T max , where N is the threshold number of data points in each bin.

[0129] 4. Transform the second tensor T along the first dimension (i.e., the modified data point coordinates) max Concatenate with the first tensor T to generate a tensor with size (2C K ,N,P)’s third tensor T’.

[0130] 5. Set the current 3D feature tensor to the third tensor T'.

[0131] After performing K iterations, the encoder 560 obtains a learned feature output 562, which is a (C, P) tensor including P feature vectors, each of which has a size of C.

[0132] The 2D image creation component 564 is configured to receive the learned feature output 562 from the encoder 560 and process the learned feature output 562 to generate a pseudo image 566. The pseudo image 566 is a 2D image with more channels (e.g., 32, 64, or 128 channels) than a standard RGB image with 3 channels.

[0133] In particular, if Figure 7 As shown, the 2D image creation component 564 uses the P-dimensional column index vector 559 to spread the dense (C, P) tensor to multiple locations on the pseudo image 566. That is, for each feature vector of size C in the dense (C, P) tensor, the image creation component 564 uses the P-dimensional column index vector 559 to find the 2D coordinates of the feature vector and places the feature vector into the pseudo image 566 at the 2D coordinates. As a result, each location on the pseudo image 566 corresponds to one of the columns and represents the features of the data points in that column.

[0134] By converting the sparse point cloud 553 into a dense pseudo image 566 that is compatible with a standard 2D convolutional architecture, the perception system 402 can efficiently and quickly process the pseudo image by leveraging the processing power and rate of a convolutional neural network (CNN) and a GPU.

[0135] Figure 5C Example point clouds and columns are illustrated. In some embodiments, the columns are polar coordinate columns. Point cloud 570 has multiple data points 572. In this embodiment, each data point is a 5-dimensional data point with a spatial location (x, y, z), reflectivity (r), and a timestamp (t). The timestamp allows multiple LiDAR or Radar scans to accumulate data points as input for a single prediction / detection of an object. Each data point is assigned to one of B = H x W columns. Each column is a z column extending from a portion of a 2D ground plane 574 in the z direction. As described above with reference to Figure 5B As described, the column processing module is configured to transform each data point having an original representation into a corresponding modified data point having a different representation. Figure 5CAs shown, the data point 576 in the non-empty column 582 has an original 5-dimensional representation including 3D spatial coordinates (x, y, z), reflectivity (r), and time stamp (t). The column processing module transforms the data point 576 represented by (x, y, z, r, t) into a data point represented by (x offset ,y offset ,z,r,t,d) represents the corresponding modified data point, where x offset and offset It is measured based on the relative distance between the data point 576 and the center 580 of the non-empty column 582, where z is the height of the data point in the non-empty column, r is the reflectivity, t is the timestamp, and d is the cylindrical Euclidean distance from the sensor 571 to the data point 576.

[0136] In an embodiment, the 2D ground plane 574 is divided into a plurality of grid cells having the same size, and thus a plurality of pillars extending vertically (in the Z direction) from these grid cells also have the same volume. However, in other embodiments, such as Figure 5D As shown, the 2D ground plane can be divided into multiple grid cells with different sizes.

[0137] Figure 5D Example 2D ground plane illustrating a point cloud. The 2D ground plane is divided into multiple grid cells with different cell sizes depending on the density of objects in the 3D space.

[0138] like Figure 5D As shown, region 592 has a high density of detected objects based on another sensor of the vehicle (e.g., camera, Radar, sonar). Therefore, region 592 has a smaller grid cell size, which means that there are more columns in region 592 to capture more information about the objects in region 592. In an embodiment, each column in region 592 has more data points allowed in each column compared to other regions with lower object density. In contrast, region 596 does not have any objects detected by the camera. Thus, there are no columns in region 596. Region 594 has a medium density of detected objects and therefore has more columns than region 596 but fewer columns than region 592. Typically, threshold N is larger for columns located in regions with high object density and smaller for columns located in regions with low object density.

[0139] Instance Segmentation System

[0140] Return to Figure 5AAs a component of instance segmentation system 505, image feature extraction network 506 can receive a pseudo image and extract features from the pseudo image. In some cases, image feature extraction network 506 is configured to process the pseudo image to generate an intermediate output of features that characterize the pseudo image. In some cases, image feature extraction network 506 is a 2D CNN including one or more neural network layers (e.g., FIG. 4A to FIG. 4D 420 of CNN 420). The one or more neural network layers may include one or more of (i) a 3×3 convolutional neural network layer, (ii) a rectified linear unit (ReLU) neural network layer, and (iii) a batch normalization neural network layer. In an embodiment, the intermediate output is a feature map with more channels than the pseudo image 566. For example, if the pseudo image has 32 channels, the intermediate output may have 512 channels. As another example, if the pseudo image 566 has 32 channels, the intermediate output may have 256 channels.

[0141] The semantic segmentation system 508 can use the output of the image feature extraction network 506 to label pixels and classify objects in the pseudo image. In some cases, the semantic segmentation system 508 can be implemented as a CNN (e.g., with FIG. 4A to FIG. 4D 420 ). In some cases, the semantic segmentation system 508 is implemented as a feed-forward convolutional neural network that, taking into account the output from the semantic segmentation system 508, can generate a classification score for the presence of an object class (e.g., a car, a pedestrian, or a bicycle) in the pseudo image. In some cases, these labels can be associated (or embedded) with pixels of the pseudo image. The higher the classification score, the more likely the corresponding object class is to exist (or the more likely the corresponding pixel is to correspond to an object of that class). In some cases, the semantic segmentation system can perform a column-level K-class classification with an output size of H×W×K (which can be the same spatial size as the input column). In some cases, all points inside a particular column share the same predicted semantic class. In some cases, the semantic segmentation system 508 assigns object classification to part or all of the segments of the pseudo image. Object classification can include, but is not limited to, intelligent object classification (e.g., vehicles, pedestrians, bicycles, etc.) and non-intelligent object classification (e.g., sidewalks, roads, curbs, etc.).

[0142] The instance part segmentation system 510 can use the output of the image feature extraction network 506 to classify the segments of the pseudo image corresponding to different parts of the instance of the object in the pseudo image. For example, the instance part segmentation system 510 can classify the segments of the pseudo image corresponding to the same instance of the agent (e.g., an object that can move and / or perform an action) as primary segments or secondary segments. In some cases, the instance part segmentation system 510 can be implemented as the head of a neural network that is trained to assign instance part classifications (scores) to the segments of the pseudo image corresponding to different parts of the agent in the pseudo image. The higher the classification score, the higher the confidence that the classification of the segment is accurate.

[0143] In some cases, ground truth data used to train a neural network associated with instance part segmentation system 510 may include labels for identifying pseudo image segments corresponding to different parts of an instance of an agent. In some cases, pseudo image segments of the ground truth data corresponding to different parts of an instance of an agent may be labeled differently to enable the neural network associated with instance part segmentation system 510 to learn to distinguish between different parts of an instance of an agent.

[0144] In some cases, the ground truth data may include labels that identify segments corresponding to primary portions of an agent (non-limiting examples: bottom-most, top-most, right-most, left-most portions of an agent) in one way, and segments corresponding to other (secondary) portions of the agent in a different way. For example, at least one pseudo-image segment corresponding to an agent may be labeled as a primary segment of the instance, and some or all other pseudo-image segments corresponding to the same agent instance may be labeled as secondary segments of the same agent instance.

[0145] In some such cases, the ground truth data may omit or not include labels for objects that do not move or act (also referred to as non-agents, e.g., sidewalks, streets, grass, curbs, etc.), and / or labels for empty segments of the pseudo image (whether or not the empty segments correspond to agents). In this way, the neural network associated with the instance part segmentation system 510 can learn to classify parts of agent objects and ignore non-agent objects.

[0146] In some cases, the instance part segmentation system 510 may assign an instance part classification (or score) to a pseudo image segment corresponding to an agent (and may ignore segments corresponding to non-agent objects and / or empty segments corresponding to agents or non-agents). In some cases, the instance part segmentation system 510 may also indicate a probability or likelihood that the instance part classification is correct.

[0147] In some cases, the instance part segmentation system 510 ignores segments corresponding to non-agent objects and / or empty segments based on ground truth data used to train a neural network associated with the instance part segmentation system 510. For example, if the ground truth data does not include labels for non-agent objects and / or empty segments, the neural network may not learn to classify such segments.

[0148] The instance part segmentation system 510 can assign the instance part classification to the segment based on the portion of the agent instance corresponding to the pseudo image segment. In some cases, the instance part segmentation system 510 can assign different instance part classifications to the segments of the pseudo image corresponding to different parts of the agent. For example, the instance part segmentation system 510 can assign one instance part classification to the segment corresponding to a specific part of the agent in the pseudo image (e.g., the main segment of the instance and the main part of the instance), and assign different instance part classifications to the segments corresponding to different parts of the same agent (e.g., the secondary segment of the instance and the secondary part of the instance). In this way, the pseudo image segment corresponding to the main part of the agent can have a different instance part classification from some or all other pseudo image segments (e.g., secondary segments) corresponding to other (secondary) parts of the same agent.

[0149] It will be understood that the primary segments of an instance may correspond to any portion of an agent as desired. In some cases, the primary segments may correspond to pseudo-image segments that include points on the perimeter, edges, or corners of an agent. For example, the primary segments may correspond to pseudo-image segments of an agent that include: a left edge or leftmost point, a right edge or rightmost point; a top edge or topmost point; a bottom edge or bottommost point; a lower left edge, corner, or point; a lower right edge, corner, or point; an upper left edge, corner, or point; an upper right edge, corner, or point; and so on.

[0150] In some cases, the instance part segmentation system 510 may classify multiple segments as primary segments. As a non-limiting example, the instance part segmentation system 510 may classify a pseudo image segment including part or all of the lower left corner, lower right corner, upper left corner, and upper right corner of the object differently from a segment including other parts of the object.

[0151] In some cases, the instance part classification for a segment associated with an agent can be binary. In some cases, the first value in the binary format can indicate that a particular segment of the agent is a primary segment and corresponds to (or includes) the primary part of the agent (e.g., the leftmost part of the agent). The second value for the instance part classification can indicate that the particular segment is a secondary segment and corresponds to (or includes) other (secondary) parts of the agent, or does not correspond to (e.g., does not include) the primary part of the agent. As described herein, in some cases, the instance part segmentation system 510 may not assign an instance part classification to a segment, which may indicate that the segment corresponds to a non-agent object or is empty (e.g., has almost no data points in itself).

[0152] In some cases, the example part segmentation system 510 can use the overall affinity vector to represent the affinity between the segments. In the case where the segments of the pseudo image correspond to columns, the columns pl i and pl j The affinity between can be expressed as a pli,plj , and with column pl i The associated overall affinity vector can be A = {a pli,plj}, j = 0, 1, ..., HW-1. In the example, the partial segmentation system 510 determines the column pl in a zigzag traversal order. i If it shares the same instance with any of its previous bars, for pl i The goal can be reduced to a' i =max({a pli,plj}), j = 0, 1, ..., i-1, the target can be a single value. i The ground truth value of can be a binary value. In some cases, a' i = 0 can mean pl i is not similar to any of its previous bars, and a' i =1 can mean pl i shares at least one instance with its previous column. In addition, depending on the direction of traversal, a' i = 0 can indicate pl i is the leftmost column of a new (or unrecognized) instance.

[0153] Instance identifier system 512 can use the output of semantic segmentation system 508 and instance part segmentation system 510 to cluster segments (and instance parts) of the pseudo image and identify different instances of agents and / or objects within the pseudo image (e.g., using the clustered segments). In some cases, instance identifier system 512 can be implemented using one or more hardware processors, etc.

[0154] In some cases, the instance identifier system 512 may scan the pseudo image in a predetermined pattern to identify the agent instance. It will be appreciated that various predetermined patterns may be used to scan the pseudo image, such as, but not limited to, zigzag (e.g., scanning a column from top to bottom, moving to the next column and scanning the column from top to bottom), back and forth (e.g., scanning a column from top to bottom, moving to the next column and scanning the column from bottom to top, etc.), etc. Similarly, various directions may be used, such as column by column (e.g., left to right, right to left) or row by row (e.g., bottom to top, top to bottom).

[0155] In some cases, the pattern and direction can correspond to the manner in which the main section / main portion of an instance is marked. For example, if the main portion corresponds to the leftmost portion of the instance, the instance identifier system 512 can use a zigzag pattern, scanning columns from left to right, wherein each column is scanned from bottom to top before moving to the next column. As another example, if the main portion corresponds to the bottommost portion of the instance, the instance identifier system 512 can use a zigzag pattern, scanning rows from bottom to top, wherein each row is scanned from right to left before moving to the next row. In addition, depending on the manner in which portions of the instance are labeled and / or classified, other patterns for traversing the pseudo-image can be used.

[0156] In some cases, when the instance identifier system 512 scans segments of a pseudo image, the instance identifier system 512 may ignore segments of the pseudo image that are not classified by the instance identifier system 512 (e.g., empty segments and / or non-agent segments) and / or segments classified by the semantic segmentation system 508 as non-agent objects.

[0157] When instance identifier system 512 encounters a segment classified by instance portion segmentation system 510 (and / or classified as corresponding to an agent object), instance identifier system 512 can determine whether the segment corresponds to an existing instance or a new (or unrecognized) instance.

[0158] In some cases, instance identifier system 512 can use the instance part classification assigned to a segment by instance part segmentation system 510 to determine whether the segment corresponds to an existing instance or a new (or unrecognized) instance. In some cases, if the instance part classification assigned to the segment identifies the segment as a primary segment, instance identifier system 512 can determine that the segment corresponds to a new (or unrecognized) instance. Based on the determination that the segment corresponds to a new (or unrecognized) instance, instance identifier system 512 can assign a unique instance identifier (e.g., an identifier that is different from all other existing instance identifiers in the pseudo image) to the new (or unrecognized) instance.

[0159] In some cases, if the instance partial classification assigned to the segment identifies the segment as a secondary segment, the instance identifier system 512 can determine that the segment corresponds to an existing instance. Based on the determination that the segment corresponds to an existing instance, the instance identifier system 512 can identify the existing instance to associate with the segment. In some cases, the instance identifier system 512 can identify the existing instance based on a distance relative to other segments within the pseudo image and / or an object classification of other segments within the pseudo image.

[0160] In some cases, the instance identifier system 512 may identify existing instances with the same object classification as the (secondary) segment. For example, different segments within the pseudo image may correspond to instances of different types of agents or objects. Some of these agents or objects (and corresponding segments) may have a different object classification than the segment being viewed. Therefore, the instance identifier system 512 may ignore instances (and corresponding segments) with different object classifications and identify existing instances (and corresponding segments) with the same object classification as the segment being viewed to associate with the segment being viewed.

[0161] In some cases, the instance identifier system 512 can identify an existing instance that includes a segment within a threshold distance of the segment being viewed. For example, there may be multiple instances of an agent or object within a pseudo image, however, some of the agents or objects may not be nearby (e.g., the corresponding segments of those agents or objects do not meet the threshold distance relative to the segment being viewed) and may correspond to different instances. Therefore, the instance identifier system 512 can identify an existing instance that includes a segment within a threshold distance of the segment being viewed to associate with the segment being viewed.

[0162] In some cases, the distance threshold can correspond to the calculated difference between two segments (the segment being viewed and the segment that has been associated with the instance) and / or a number of rows / columns. For example, if the instance identifier system 512 is traversing from left to right and from bottom to top in a zigzag pattern, the distance threshold can be four columns, so that a segment corresponding to an object of a first type (e.g., a vehicle) that is five columns away from a segment corresponding to an object of the same type (e.g., a vehicle) will be identified as a different instance of an object of that type (e.g., another vehicle). As another example, if the distance threshold is three segments, the instance identifier system 512 can determine whether two segments associated with the same type of agent are more than three segments apart (considering rows and columns). It will be understood that other units can be used to calculate distances, such as mm, cm, m, etc.

[0163] In some cases, when determining whether the distance threshold is met, instance identifier system 512 can take into account that the opposing edges of the pseudo image can correspond to adjacent locations in the scene. For example, if the pseudo image corresponds to a 360-degree view of the scene, the leftmost segment and the rightmost segment of the pseudo image can correspond to adjacent locations in the scene. In such a scenario, instance identifier system 512 can calculate the distance between the segments as if the leftmost segment and the rightmost segment of the pseudo image were adjacent.

[0164] In some cases, the threshold distance corresponds to an instance having a segment that is closest to the segment being viewed and has the same object classification as the segment being viewed. For example, there may be multiple instances of agents that are near a segment and have the same object classification as the segment. Therefore, in some cases, the instance identifier system 512 can identify an existing instance (of the same object classification) that includes the segment that is closest to the segment to associate with the segment being viewed.

[0165] Once an existing instance is identified for association, instance identifier system 512 can associate the segment being viewed with the identified instance and with other segments corresponding to the instance. In some cases, instance identifier system 512 can associate the segment with the existing instance by assigning the segment an instance identifier that is the same as the instance identifier for the existing instance (e.g., the same instance identifier assigned to other segments associated with the instance). In this way, the segment being viewed can be assigned the same instance identifier as the existing instance.

[0166] In some cases, by separating semantic segmentation and classifying different parts of an instance, perception system 402 can reduce the computational requirements of instance identifier system 512. For example, by classifying the primary portion of an instance (not limited to portions at edges or corners of an instance) differently from the secondary portion of the instance, perception system 402 can reduce the complexity of identifying the instance by instance identifier system 512. In some such cases, this can reduce the computational resources used to identify instances in an image. The reduced complexity can reduce the processing time for identifying instances and improve the responsiveness of the vehicle. In addition, this can enable the vehicle to more quickly and accurately identify objects in a scene and plan routes through the scene.

[0167] Fig. 6A and Figure 6B is an operational diagram illustrating example operations of perception system 402 to identify an instance of an agent in a pseudo image.

[0168] At step 602, pseudo image system 504 generates pseudo image 652 using 3D image data 650. As described herein, 3D image data 650 may correspond to LiDAR data, Radar data, etc. In some cases, the pseudo image may be a bird's-eye view image. As described herein, the pseudo image may be a 2D image including a plurality of segments. In some cases, a segment may be a column extending in an upward direction from a ground plane. As described herein, a segment may include one or more data points associated with an object in a scene.

[0169] In some cases, the pseudo image system 504 can extract one or more learned features from a segment of an image. As described herein, the extracted features can be learned features of different segments of an image.

[0170] In the illustrated example, the pseudo image 652 includes six columns of segments (or columns) by six rows of segments (or columns), however, it will be appreciated that the pseudo image 652 may include fewer or more segments as desired. In some cases, each segment may correspond to a specific location of the scene of the vehicle. As described herein, in some cases, a segment may correspond to a 20 cm by 20 cm portion of the scene of the vehicle.

[0171] At step 604, the image feature extraction network 506 extracts one or more features from the pseudo image 652. As described herein, the extracted features may be learned features of one or more objects within the pseudo image. In this way, the image feature extraction network may provide an enriched pseudo image. For example, a portion or all of the data points within the pseudo image may be further associated with features extracted from the pseudo image by the image feature extraction network 506. In some cases, the extracted features may be embedded with data points and / or segments of the pseudo image. Image 654 illustrates an example of an enriched pseudo image.

[0172] At step 606, the semantic segmentation system 508 segments the segments of the pseudo image into one or more object classifications. As described herein, the semantic segmentation system 508 can classify the segments within the pseudo image into any one or any combination of the following: obstacles, bicycles, buses, cars, construction vehicles, motorcycles, pedestrians, traffic cones, trailers, trucks, drivable surfaces, other flat surfaces, sidewalks, terrain, man-made structures, vegetation. In some cases, one or more of the classifications may correspond to intelligent agents (e.g., bicycles, buses, cars, construction vehicles, motorcycles, pedestrians, trailers, trucks), and other classifications may correspond to non-intelligent agents (e.g., obstacles, traffic cones, drivable surfaces, other flat surfaces, sidewalks, terrain, man-made structures, vegetation). In some cases, such as the pseudo image 656 ( Figure 6B), object classification can be embedded in a pseudo image with different sections, where different patterns 657A, 657B, 657C, 657D, 657E indicate different object classifications for each section. In the illustrated example, patterns 657A and 657B correspond to agent object classifications, and patterns 657C, 657D, 657E correspond to non-agent object classifications.

[0173] In step 608, the instance part segmentation system 510 classifies the segments based on the corresponding parts of the instance. As described herein, the instance part segmentation system 510 can classify the segments based on the parts of the agent instance associated with the segments of the pseudo image. For example, if the segment corresponds to the main part of the instance of the agent, the instance part segmentation system 510 can classify the segment as a main segment. If the segment does not correspond to the main part of the instance of the agent (or corresponds to the secondary part), the instance part segmentation system 510 can classify the segment as a secondary segment. In some cases, the instance part segmentation system 510 may not classify the segment. For example, the instance part segmentation system 510 may not classify the segment. For example, the instance part segmentation system 510 may not classify the empty (or having less than a threshold amount of data points) segments, and / or may not classify the segments that do not correspond to the agent (e.g., the segments corresponding to the non-agent object).

[0174] In some cases, the instance part segmentation system 510 can classify the segments by assigning instance part classification scores. Pseudo image 658 illustrates an example of a pseudo image in which some segments are assigned instance part classification scores. In the illustrated example, the instance part classification "0" indicates a segment classified as a primary segment, and the instance part classification "1" indicates a segment classified as a secondary segment. As described herein, the instance part segmentation system 510 can classify segments corresponding to intelligent objects, and can ignore segments corresponding to non-intelligent objects and / or that do not meet a threshold amount of data points (e.g., empty segments), or not classify such segments.

[0175] At step 610, instance identifier system 512 combines features from semantic segmentation system 508 and instance part segmentation system 510. In some cases, instance identifier system 512 combines features from semantic segmentation system 508 and instance part segmentation system 510 by segment, such that object classifications (or other features) determined by semantic segmentation system 508 are aggregated to instance part classifications (or other features) determined for the same segment by instance part segmentation system 510. Pseudo image 660 illustrates this by showing class identifications (e.g., patterns 657A-657E) aggregated with instance part classification scores for the relevant segment.

[0176] At step 612, the instance identifier system 512 clusters (or groups) segments associated with the same instance. As described herein, as shown in image 662, the instance identifier system 512 can traverse the pseudo image in a predetermined pattern. In the illustrated example, the instance identifier system 512 traverses the columns of the pseudo image 662 in a zigzag pattern from left to right (moving from the bottom of each column to the top). When the instance identifier system 512 encounters an instance part classification, it determines whether the instance part classification corresponds to a new (unrecognized) instance (e.g., "0") or an existing instance (e.g., "1"). In the illustrated example, by traversing the image 662 according to the predetermined pattern, the instance identifier system 512 can be sure that the leftmost portion of the instance is the first portion encountered. In this way, the instance identifier system 512 can effectively use the leftmost portion of the instance of the object to distinguish different instances. In this way, the instance identifier system 512 can reduce the computational complexity of grouping or clustering segments.

[0177] If the instance partial classification indicates a new (or unidentified) instance, the instance identifier system 512 generates a new (unique) instance identifier for the new (or unidentified) instance (e.g., A1, A2, B1). In some cases, the unique identifier may also indicate the class or type of the instance.

[0178] If the instance partial classification indicates an association with an existing instance, the instance identifier system 512 identifies the existing instance to which the segment can be associated, and associates the segment with the instance by assigning the segment the same instance identifier as other segments of the existing instance. In some cases, the instance identifier system 512 can identify the existing instance to which the segment can be associated based on an object classification associated with the segment and a distance threshold. For example, as described herein, the instance identifier system 512 can identify an existing instance that has the same object classification as the segment being viewed and satisfies a distance threshold (e.g., a segment of the existing instance satisfies the distance threshold relative to the segment being viewed).

[0179] Pseudo image 656 illustrates an example result of clustering segments, where "A1" and "A2" indicate different instances of objects of the same type or classification (e.g., vehicles) as indicated by shared pattern 657A, and "B1" indicates instances of objects of different types or classifications (e.g., pedestrians) as indicated by different pattern 657B. Segments with patterns 657C to 657E may indicate segments corresponding to non-intelligent objects.

[0180] As described herein, segments associated with the same instance of an object may share the same object classification and satisfy a distance threshold. Thus, even though the "B1" segments are not adjacent, instance identifier system 512 may group them together.

[0181] Furthermore, in the illustrated example, instance identifier system 512 has identified the segments on the rightmost column and the leftmost column as being part of the same "A1" instance. As described herein, instance identifier system 512 can determine that the rightmost portion and the leftmost portion of the pseudo image correspond to adjacent locations of the vehicle scene. Thus, instance identifier system 512 can determine that those portions satisfy the distance threshold regardless of their location on pseudo image 656.

[0182] As instance identifier system 512 traverses pseudo image 662, it can ignore segments that do not meet a threshold amount of data points (e.g., empty segments) and / or segments associated with non-agents. In some cases, instance identifier system 512 passively ignores these segments by clustering segments that have instance part classifications and ignoring segments that do not have instance part classifications.

[0183] Example flow chart of a test system

[0184] Figure 7 is a flow chart illustrating an example of a routine 700 for segmenting an instance of an object implemented by one or more processors. Figure 7 The illustrated flowchart is provided for illustration purposes only. It will be understood that the Figure 7 One or more of the steps of the illustrated routines, or the order of the steps may be changed. In addition, for the purpose of illustrating clear examples, one or more specific system components are described in the context of performing various operations during various data flow stages. However, other system arrangements and distribution of processing steps across system components may be used.

[0185] At box 702, the instance segmentation system 505 receives a pseudo image. As described herein, the pseudo image can be generated by the pseudo image system 504 based on a 3D image received from a 3D image sensor attached to a vehicle (such as, but not limited to, a LiDAR sensor, a Radar sensor, etc.). The pseudo image may include a 2D representation of the 3D image and may include multiple segments (or columns). Some or all of the segments may include one or more data points, which include spatial coordinates (e.g., x, y, z) and other information, such as, but not limited to, offset, reflectivity, time, distance from a specific point to a specific location within the pseudo image segment (e.g., the center of the pseudo image segment), distance from a specific point to the image sensor that detected the point, etc. As described herein, each data point may include fewer or more dimensions or features.

[0186] At block 704, instance segmentation system 505 determines an object classification for the segment of the pseudo image. As described herein, in some cases, instance segmentation system 505 may classify objects within the image using one or more neural networks associated with semantic segmentation system 508. In some cases, objects may be classified as intelligent entities (e.g., bicycles, buses, cars, construction vehicles, motorcycles, pedestrians, trailers, trucks) or non-intelligent entities (e.g., obstacles, traffic cones, drivable surfaces, other flat surfaces, sidewalks, terrain, man-made structures, vegetation).

[0187] A pseudo image segment that includes a portion of an object may be assigned the same classification as the object. Thus, if a segment includes a bicycle object, the segment may be assigned the object classification: bicycle, etc. In some cases, the pseudo image system 504 may generate a pseudo image to avoid overlapping objects in a segment. For example, the pseudo image may be generated as a 2D bird's-eye view to reduce or eliminate the likelihood that a segment includes more than one type of object.

[0188] At block 706, the instance segmentation system 505 determines an instance part classification for the segment. As described herein, in some cases, the segmentation system 505 may classify parts of instances within the image using one or more neural networks associated with the instance part segmentation system 510. In some cases, the instance segmentation system 505 classifies (only) parts of agent objects and may not classify non-agent objects, however, it will be appreciated that the instance segmentation system 505 may be configured in a variety of ways.

[0189] In some cases, the instance segmentation system 505 may classify a portion of an instance as a primary portion or a secondary portion. In some cases, as described herein, if a portion of an instance corresponds to a specific (or predetermined) portion of an instance (such as, but not limited to, the leftmost, rightmost, topmost, or bottommost portion of an instance, a location on the periphery or edge of an instance, etc.), the instance segmentation system 505 may classify the portion as a primary portion of the instance. Similarly, if a portion of an instance corresponds to another portion of the instance (e.g., a portion other than the primary portion), the instance segmentation system 505 may classify the portion as a secondary portion of the instance. In some cases, an instance may include a single primary portion, in which case the rest of the instance may be a secondary portion. In some cases, an instance may include multiple primary portions, in which case the secondary portion may include a portion other than the primary portion.

[0190] A pseudo image segment that includes a portion of an intelligent object can be assigned the same instance part classification as the portion of the intelligent object. Thus, if the segment includes a primary portion of the intelligent object, the instance segmentation system 505 can identify the segment as a primary segment. If the segment includes a secondary portion of the intelligent object, the instance segmentation system 505 can identify the segment as a secondary segment.

[0191] At block 708, the instance segmentation system 505 groups the segment with one or more other segments of the pseudo image. As described herein, to group the segments, the instance segmentation system 505 can traverse the image. In some cases, the instance segmentation system 505 traverses the pseudo image in a predetermined pattern, such as from bottom to top and from left to right in a column. By traversing the image in a predetermined pattern, the instance segmentation system 505 can reduce the amount of computing resources used to identify instances.

[0192] In some cases, instance segmentation system 505 can group segments with other segments based on the segment's object classification and / or the segment's instance part classification. In some cases, instance segmentation system 505 can group segments with other segments that have the same object classification. For example, if a segment has an object classification of "vehicle," instance segmentation system 505 can limit the segments that can be grouped with the segment being viewed to those that also have an object classification of "vehicle."

[0193] In some cases, the instance segmentation system 505 can group the segment being viewed with other segments based on the instance portion classification of the segment. For example, if the instance portion classification for a segment identifies the segment as a primary segment, or the instance segmentation system 505 otherwise determines that the segment corresponds to a new or unrecognized instance, the instance segmentation system 505 can assign a new instance identifier to the segment. As the instance segmentation system 505 continues to traverse the pseudo image and encounters secondary segments that have the same object classification as the segment and meet the distance threshold, the instance segmentation system 505 can assign the same instance identifier to these secondary segments.

[0194] As another example, if the instance partial classification for a segment identifies the segment as a secondary segment, or otherwise determines that the segment corresponds to an existing instance, the instance segmentation system 505 can identify an existing instance having the same object classification and satisfying a threshold distance (e.g., including a segment satisfying a threshold distance from the segment being viewed). The instance segmentation system 505 can associate an instance identifier of the identified instance with the segment being viewed.

[0195] At block 710, instance segmentation system 505 causes the vehicle to be operated based on the groupings. For example, instance segmentation system 505 may communicate the groupings to another component of the perception system and / or to planning system 404. Based on the groupings and the identified instances of objects within the scene, planning system 404 may plan a route for the vehicle and control the vehicle using control system 408.

[0196] It will be understood that fewer, more, or different boxes may be used in routine 700. In addition, one or more boxes may be rearranged or performed simultaneously or in parallel. In some cases, one or more boxes (e.g., boxes 704 to 708) of routine 700 may be repeated multiple times for multiple segments of the pseudo image until a portion or all of the instances of the object in the image are identified and classified. Thus, in some cases, instance segmentation system 505 may identify multiple primary segments (as described herein) and multiple secondary segments (as described herein) within the pseudo image, and then associate the secondary segments with one of the primary segments (e.g., based on object classification and distance). By grouping the secondary segments with the primary segments, instance segmentation system 505 may identify multiple instances of the object. Autonomous vehicle computing 400 may control the vehicle using the identified instances of the object, etc.

[0197] As another example, in some cases, the routine 700 may include processing the pseudo image having the grouped segments using one or more prediction systems, and planning a route based on outputs of the prediction systems.

[0198] In the previous description, aspects and embodiments of the present disclosure have been described with reference to many specific details, which may vary depending on the implementation. Therefore, the description and the accompanying drawings should be regarded as illustrative, not restrictive. The only and exclusive indication of the scope of the invention, and the applicant's expectation that the content of the scope of the invention is the literal and equivalent scope of the claims issued from this application in the specific form of the claims, including any subsequent amendments. Any definition of the terms used to be included in such claims that are clearly set forth herein should be based on the meaning of such terms as used in the claims. In addition, when the term "also includes" is used in the previous description or the attached claims, the following of the phrase may be an additional step or entity, or a sub-step / sub-entity of the previously described step or entity.

[0199] Various additional example embodiments of the present disclosure may be described by the following clauses:

[0200] Clause 1. A method comprising:

[0201] receiving, with at least one processor, a pseudo image of a 3D image, wherein the 3D image is received from a 3D image sensor coupled to a vehicle, and wherein the pseudo image includes a plurality of segments;

[0202] determining, with the at least one processor, an object classification for a segment of a plurality of segments of the pseudo image;

[0203] determining, with the at least one processor, an instance portion classification for the segment of the pseudo image;

[0204] grouping the segment with another segment based at least in part on the object classification of the segment and the instance portion classification for the segment; and

[0205] The vehicle is operated based at least in part on grouping the segment with the other segment.

[0206] Clause 2. The method of clause 1, wherein at least one segment of the plurality of segments comprises a plurality of 3D data points.

[0207] Clause 3. The method of clause 1 or 2, wherein the pseudo image comprises a 2D representation of the 3D image.

[0208] Clause 4. The method of any one of clauses 1 to 3, wherein determining the object classification for at least one segment comprises:

[0209] identifying at least a portion of an object in the segment;

[0210] determining an object classification for at least a portion of the objects in the segment; and

[0211] An object classification for at least a portion of the object is associated with the segment.

[0212] Clause 5. The method of any one of clauses 1 to 4, wherein the object classification for at least one segment includes one of a vehicle, a pedestrian, and a bicycle.

[0213] Clause 6. The method of any one of clauses 1 to 5, wherein determining the instance partial classification for at least one segment comprises:

[0214] identifying at least a portion of instances of Agent objects in the segment;

[0215] determining an instance part classification for at least a portion of the agent objects in the segment; and

[0216] An instance portion classification for at least a portion of the agent objects in the segment is associated with the segment.

[0217] Clause 7. The method of clause 6, wherein the instance partial classification for the at least one segment indicates that the segment is one of a primary segment of the instance of the agent object and a secondary segment of the instance of the agent object.

[0218] Clause 8. A method according to clause 6, wherein determining the instance portion classification for the at least one segment includes: based on determining that the segment includes a predetermined portion of the instance of the intelligent object, identifying the segment as a primary segment of the instance of the intelligent object.

[0219] Clause 9. The method of clause 8, wherein the predetermined portion of the instance of the agent object is a leftmost portion of the instance of the agent object.

[0220] Clause 10. A method according to clause 6, wherein determining the instance portion classification for the at least one segment includes: based on determining that the segment does not include a predetermined portion of the instance of the intelligent object, identifying the segment as a secondary segment of the instance of the intelligent object.

[0221] Clause 11. The method of any one of clauses 1 to 7, wherein grouping at least one segment with another segment comprises:

[0222] traversing the pseudo image according to a predetermined pattern;

[0223] determining that the instance partial classification for the at least one segment indicates that the at least one segment corresponds to an unidentified instance;

[0224] generating an instance identifier based on determining that the partial classification of the instance for the at least one segment corresponds to an unidentified instance;

[0225] associating the instance identifier with the at least one segment;

[0226] traversing the pseudo image according to the predetermined pattern;

[0227] determining that the partial classification of the instance for the another segment corresponds to an existing instance;

[0228] determining that the object classification for the at least one segment matches the object classification for the another segment;

[0229] determining that a distance between the at least one segment and the another segment satisfies a threshold distance; and

[0230] The generated instance identifier is associated with the other segment.

[0231] Clause 12. The method of any one of clauses 1 to 7, wherein grouping at least one segment with another segment comprises:

[0232] traversing the pseudo image according to a predetermined pattern;

[0233] determining that the instance partial classification for the at least one segment corresponds to an existing instance; and

[0234] Based on determining that the instance portion classification for the at least one segment corresponds to an existing instance:

[0235] identifying a previously traversed section of the pseudo image,

[0236] determining that the object classification for the at least one segment matches the object classification for the previously traversed segment,

[0237] determining that a distance between the at least one segment and the previously traversed segment satisfies a threshold distance, and

[0238] An instance identifier associated with the previously traversed segment is associated with the at least one segment.

[0239] Clause 13. A method according to any one of clauses 1 to 7, 11 and 12, wherein the instance part classification is binary.

[0240] Clause 14. A system comprising:

[0241] at least one processor; and

[0242] at least one non-transitory storage medium storing instructions that, when executed by the at least one processor, cause the at least one processor to:

[0243] receiving a pseudo image of a 3D image, wherein the 3D image is received from a 3D image sensor coupled to a vehicle, and wherein the pseudo image includes a plurality of segments;

[0244] determining an object classification for a segment of a plurality of segments of the pseudo image;

[0245] determining an instance portion classification for the segment of the pseudo image;

[0246] grouping the segment with another segment based at least in part on the object classification of the segment and the instance portion classification for the segment; and

[0247] The vehicle is caused to be operated based at least in part on grouping the segment with the other segment.

[0248] Clause 15. The system of clause 14, wherein, to group at least one section with another section, the instructions cause the at least one processor to:

[0249] traversing the pseudo image according to a predetermined pattern;

[0250] determining that the instance partial classification for the at least one segment indicates that the at least one segment corresponds to an unidentified instance;

[0251] generating an instance identifier based on determining that the partial classification of the instance for the at least one segment corresponds to an unidentified instance;

[0252] associating the instance identifier with the at least one segment;

[0253] traversing the pseudo image according to the predetermined pattern;

[0254] determining that the partial classification of the instance for the another segment corresponds to an existing instance;

[0255] determining that the object classification for the at least one segment matches the object classification for the another segment;

[0256] determining that a distance between the at least one segment and the another segment satisfies a threshold distance; and

[0257] The generated instance identifier is associated with the other segment.

[0258] Clause 16. The system of clause 14, wherein, to group at least one section with another section, the instructions cause the at least one processor to:

[0259] traversing the pseudo image according to a predetermined pattern;

[0260] determining that the instance partial classification for the at least one segment corresponds to an existing instance; and

[0261] Based on determining that the instance portion classification for the at least one segment corresponds to an existing instance:

[0262] identifying a previously traversed section of the pseudo image,

[0263] determining that the object classification for the at least one segment matches the object classification for the previously traversed segment,

[0264] determining that a distance between the at least one segment and the previously traversed segment satisfies a threshold distance, and

[0265] An instance identifier associated with the previously traversed segment is associated with the at least one segment.

[0266] Clause 17. The system of any one of clauses 14 to 16, wherein, to determine an instance partial classification for at least one segment, the instructions cause the at least one processor to:

[0267] identifying at least a portion of instances of Agent objects in the segment;

[0268] determining an instance part classification for at least a portion of the agent objects in the segment; and

[0269] An instance portion classification for at least a portion of the agent objects in the segment is associated with the segment.

[0270] Clause 18. At least one non-transitory storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to:

[0271] receiving a pseudo image of a 3D image, wherein the 3D image is received from a 3D image sensor coupled to a vehicle, and wherein the pseudo image includes a plurality of segments;

[0272] determining an object classification for a segment of a plurality of segments of the pseudo image;

[0273] determining an instance portion classification for the segment of the pseudo image;

[0274] grouping the segment with another segment based at least in part on the object classification of the segment and the instance portion classification for the segment; and

[0275] The vehicle is caused to be operated based at least in part on grouping the segment with the other segment.

[0276] Clause 19. At least one non-transitory storage medium according to clause 18, wherein, to group at least one section with another section, the instructions cause the at least one processor to:

[0277] traversing the pseudo image according to a predetermined pattern;

[0278] determining that the instance partial classification for the at least one segment indicates that the at least one segment corresponds to an unidentified instance;

[0279] generating an instance identifier based on determining that the partial classification of the instance for the at least one segment corresponds to an unidentified instance;

[0280] associating the instance identifier with the at least one segment;

[0281] traversing the pseudo image according to the predetermined pattern;

[0282] determining that the partial classification of the instance for the another segment corresponds to an existing instance;

[0283] determining that the object classification for the at least one segment matches the object classification for the another segment;

[0284] determining that a distance between the at least one segment and the another segment satisfies a threshold distance; and

[0285] The generated instance identifier is associated with the other segment.

[0286] Clause 20. At least one non-transitory storage medium according to clause 18, wherein, to group at least one section with another section, the instructions cause the at least one processor to:

[0287] traversing the pseudo image according to a predetermined pattern;

[0288] determining that the instance partial classification for the at least one segment corresponds to an existing instance; and

[0289] Based on determining that the instance portion classification for the at least one segment corresponds to an existing instance:

[0290] identifying a previously traversed section of the pseudo image,

[0291] determining that the object classification for the at least one segment matches the object classification for the previously traversed segment,

[0292] determining that a distance between the at least one segment and the previously traversed segment satisfies a threshold distance, and

[0293] An instance identifier associated with the previously traversed segment is associated with the at least one segment.

Claims

1. A method comprising: receiving, with at least one processor, a pseudo image of a 3D image, wherein the 3D image is received from a 3D image sensor coupled to a vehicle, and wherein the pseudo image includes a plurality of segments; determining, with the at least one processor, an object classification for a segment of a plurality of segments of the pseudo image; determining, with the at least one processor, an instance portion classification for the segment of the pseudo image; grouping the segment with another segment based at least in part on the object classification of the segment and the instance portion classification for the segment; and The vehicle is operated based at least in part on grouping the segment with the other segment.

2. The method according to claim 1, wherein: At least one segment of the plurality of segments includes a plurality of 3D data points.

3. The method according to claim 1 or 2, wherein: The pseudo image comprises a 2D representation of the 3D image.

4. The method according to any one of claims 1 to 3, wherein: Determining a classification of the object for at least one segment includes: identifying at least a portion of an object in the segment; determining an object classification for at least a portion of the objects in the segment; and An object classification for at least a portion of the object is associated with the segment.

5. The method according to any one of claims 1 to 4, wherein: The object classification for at least one segment includes one of a vehicle, a pedestrian, and a bicycle.

6. The method according to any one of claims 1 to 5, wherein: Determining a partial classification of the instance for at least one segment includes: identifying at least a portion of instances of Agent objects in the segment; determining an instance part classification for at least a portion of the agent objects in the segment; and An instance portion classification for at least a portion of the agent objects in the segment is associated with the segment.

7. The method according to claim 6, wherein: The instance portion classification for the at least one segment indicates that the segment is one of a primary segment of the instance of the agent object and a secondary segment of the instance of the agent object.

8. The method according to claim 6, wherein: Determining an instance portion classification for the at least one segment includes identifying the segment as a primary segment of the instance of the agent object based on determining that the segment includes a predetermined portion of the instance of the agent object.

9. The method according to claim 8, wherein: The predetermined portion of the instance of the agent object is a leftmost portion of the instance of the agent object.

10. The method according to claim 6, wherein: Determining an instance portion classification for the at least one segment includes identifying the segment as a secondary segment of the instance of the agent object based on determining that the segment does not include a predetermined portion of the instance of the agent object.

11. The method according to any one of claims 1 to 7, wherein: Grouping at least one segment with another segment includes: traversing the pseudo image according to a predetermined pattern; determining that the instance partial classification for the at least one segment indicates that the at least one segment corresponds to an unidentified instance; generating an instance identifier based on determining that the partial classification of the instance for the at least one segment corresponds to an unidentified instance; associating the instance identifier with the at least one segment; traversing the pseudo image according to the predetermined pattern; determining that the partial classification of the instance for the another segment corresponds to an existing instance; determining that the object classification for the at least one segment matches the object classification for the another segment; determining that a distance between the at least one segment and the another segment satisfies a threshold distance; and The generated instance identifier is associated with the other segment.

12. The method according to any one of claims 1 to 7, wherein: Grouping at least one segment with another segment includes: traversing the pseudo image according to a predetermined pattern; determining that the instance partial classification for the at least one segment corresponds to an existing instance; and Based on determining that the instance portion classification for the at least one segment corresponds to an existing instance: identifying a previously traversed section of the pseudo image, determining that the object classification for the at least one segment matches the object classification for the previously traversed segment, determining that a distance between the at least one segment and the previously traversed segment satisfies a threshold distance, and An instance identifier associated with the previously traversed segment is associated with the at least one segment.

13. The method according to any one of claims 1 to 7, 11 and 12, wherein: The instance part classification is binary.

14. A system comprising: at least one processor; as well as at least one non-transitory storage medium storing instructions that, when executed by the at least one processor, cause the at least one processor to: receiving a pseudo image of a 3D image, wherein the 3D image is received from a 3D image sensor coupled to a vehicle, and wherein the pseudo image includes a plurality of segments; determining an object classification for a segment of a plurality of segments of the pseudo image; determining an instance portion classification for the segment of the pseudo image; grouping the segment with another segment based at least in part on the object classification of the segment and the instance portion classification for the segment; and The vehicle is caused to be operated based at least in part on grouping the segment with the other segment.

15. The system of claim 14, wherein: To group at least one section with another section, the instructions cause the at least one processor to: traversing the pseudo image according to a predetermined pattern; determining that the instance partial classification for the at least one segment indicates that the at least one segment corresponds to an unidentified instance; generating an instance identifier based on determining that the partial classification of the instance for the at least one segment corresponds to an unidentified instance; associating the instance identifier with the at least one segment; traversing the pseudo image according to the predetermined pattern; determining that the partial classification of the instance for the another segment corresponds to an existing instance; determining that the object classification for the at least one segment matches the object classification for the another segment; determining that a distance between the at least one segment and the another segment satisfies a threshold distance; and The generated instance identifier is associated with the other segment.

16. The system of claim 14, wherein: To group at least one section with another section, the instructions cause the at least one processor to: traversing the pseudo image according to a predetermined pattern; determining that the instance portion classification for the at least one segment corresponds to an existing instance; as well as Based on determining that the instance portion classification for the at least one segment corresponds to an existing instance: identifying a previously traversed section of the pseudo image, determining that the object classification for the at least one segment matches the object classification for the previously traversed segment, determining that a distance between the at least one segment and the previously traversed segment satisfies a threshold distance, and An instance identifier associated with the previously traversed segment is associated with the at least one segment.

17. A system according to any one of claims 14 to 16, wherein: To determine an instance partial classification for at least one segment, the instructions cause the at least one processor to: identifying at least a portion of instances of Agent objects in the segment; determining an instance portion classification for at least a portion of the agent objects in the segment; as well as An instance portion classification for at least a portion of the agent objects in the segment is associated with the segment.

18. At least one non-transitory storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to: A pseudo image of a 3D image is received, wherein The 3D image is received from a 3D image sensor coupled to a vehicle, and wherein the pseudo image includes a plurality of segments; determining an object classification for a segment of a plurality of segments of the pseudo image; determining an instance portion classification for the segment of the pseudo image; grouping the segment with another segment based at least in part on the object classification of the segment and the instance portion classification for the segment; and The vehicle is caused to be operated based at least in part on grouping the segment with the other segment.

19. The at least one non-transitory storage medium of claim 18, wherein: To group at least one section with another section, the instructions cause the at least one processor to: traversing the pseudo image according to a predetermined pattern; determining that the instance partial classification for the at least one segment indicates that the at least one segment corresponds to an unidentified instance; generating an instance identifier based on determining that the partial classification of the instance for the at least one segment corresponds to an unidentified instance; associating the instance identifier with the at least one segment; traversing the pseudo image according to the predetermined pattern; determining that the partial classification of the instance for the another segment corresponds to an existing instance; determining that the object classification for the at least one segment matches the object classification for the another segment; determining that a distance between the at least one segment and the another segment satisfies a threshold distance; and The generated instance identifier is associated with the other segment.

20. The at least one non-transitory storage medium of claim 18, wherein: To group at least one section with another section, the instructions cause the at least one processor to: traversing the pseudo image according to a predetermined pattern; determining that the instance portion classification for the at least one segment corresponds to an existing instance; as well as Based on determining that the instance portion classification for the at least one segment corresponds to an existing instance: identifying a previously traversed section of the pseudo image, determining that the object classification for the at least one segment matches the object classification for the previously traversed segment, determining that a distance between the at least one segment and the previously traversed segment satisfies a threshold distance, and An instance identifier associated with the previously traversed segment is associated with the at least one segment.

Images