A machine learning system for parameterizing building information from building images.

Abstract

To provide a machine learning system for parameterizing building information using images, a method, and a computer-readable medium.SOLUTION: A method implemented by a machine learning system includes: receiving an image of a building; applying a machine learning model to the received image of a building to generate, for a new building, new building parameters to be input to a BIM (building information modeling) data generation system, the machine learning model being trained using images of buildings and corresponding building parameters for the buildings; and outputting the new building parameters for the new building.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present disclosure relates to machine learning systems, and more particularly to machine learning systems for building design.

Background Art

[0002] Machine learning (ML) generally involves training a machine learning model with training data to generate a trained model capable of generalizing the characteristics of data based on patterns similar to the training data. In training the model, often the parameters of the model are learned by optimizing an objective function, and thus optimizing the likelihood that the training data is observed for a given model. In some applications, the trained model needs to satisfy additional characteristics important for its domain in addition to minimizing the objective function.

Summary of the Invention

Problems to be Solved by the Invention

[0003] Generally, the present disclosure describes techniques for parameterizing building information using images. For example, a machine learning system can train a machine learning model with training data including an image of a building and corresponding known descriptive data for one or more building parameters of the building, such as building dimensions, number of floors, window positions and dimensions, floor height, interior layout, etc. In inference mode, the machine learning system can apply the trained model to an image of a building to predict the values of the building parameters of that building. The predicted values of those building parameters can be output for use in designing a new building or input into a BIM (building information modeling) data generation system to generate BIM data for a new building. The BIM data generation system can use the generated building parameters to reproduce, for example, the exterior facade or architectural style of the building in a three-dimensional (3D) building model.

[0004] The technology disclosed herein provides one or more technical advantages that enable one or more practical applications. In current architectural practice, architects manually determine building parameters or manually lay out each building element in design programs such as computer-aided design (CAD) programs to generate a desired building. This can be time-consuming and require specialized knowledge and skills of the program. The technology disclosed herein enables architects to quickly generate 3D architectural models with similar building parameters to existing buildings for which they own images, thereby simplifying building design. The architect can then modify the building parameters predicted from the machine learning model of the existing building to more quickly arrive at a new candidate building that meets the requirements of the architectural project and other architectural intentions. Once a new candidate building is arrived at using the generated building parameters, the BIM data generation system can process those building parameters and output BIM data for that new candidate building. Such requirements may be, for example, those that meet or are useful in meeting structural engineering, environmental planning, and legal planning requirements. [Means for solving the problem]

[0005] In one example, a machine learning system for predicting building parameters for an image of a building comprises an input device configured to receive an input including an image of a building, and a processing circuit and memory for executing the machine learning system. The machine learning system is configured to apply a machine learning model, which has been trained using an image of a building and its corresponding building parameters, to the received image of the building to generate new building parameters for the new building, which are then input to a BIM data generation system. The output device is configured to output the new building parameters for the new building.

[0006] In one example, the method comprises: receiving an image of a building; a machine learning system applying a machine learning model, which has been trained using the building image and the corresponding building parameters of the building, to the received building image to generate new building parameters for the new building that are input to the BIM data generation system; and the machine learning system outputting the new building parameters for the new building.

[0007] In one example, a non-temporary computer-readable medium includes machine-readable instructions for causing a processing circuit to perform an action. The action comprises receiving an image of a building, applying a machine learning model trained using the building image and the corresponding building parameters of the building to the received building image to generate new building parameters for the new building, which are then input to a BIM data generation system, and outputting the new building parameters for the new building.

[0008] Details of one or more examples of the technology of this disclosure are shown in the accompanying drawings and the following description. Other features, purposes, and advantages of the technology of this disclosure will become apparent from the description and drawings and the claims. [Brief explanation of the drawing]

[0009] [Figure 1] Block diagram showing an example system using the technology disclosed herein. [Figure 2] This is an example of a computing system using the technology described herein. [Figure 3] This flowchart shows an example of a mode of operation for a machine learning system using the technology described in this disclosure. [Modes for carrying out the invention]

[0010] Similar reference numerals refer to the same elements throughout the drawings and specifications.

[0011] Figure 1 is a block diagram showing an example system 100 according to the technology of this disclosure. As shown, system 100 comprises a user device 108, a computing system 101, and a BIM data generation system 130.

[0012] The computing system 101 runs the machine learning system 102. The machine learning system 102 may be implemented as software, but in some examples it may include any combination of hardware, firmware, and software. The machine learning system 102 trains the machine learning model 103. In this example, the machine learning model 103 comprises an image processing model 106 and one or more BIM data models 122A to 122N (collectively referred to as "BIM data model 122"). Each of the image processing model 106 and the BIM data model 122 may be a different machine learning model implemented by the machine learning system 102 and combined to form the overall machine learning model 103.

[0013] Computing system 101 may be implemented as any suitable computing system, such as one or more server computers, workstations, laptops, mainframes, appliances (dedicated devices), cloud computing systems, smartphones, tablet computers, and / or other computing systems capable of performing the operations and / or functions described in accordance with one or more aspects of this disclosure. In some examples, computing system 101 is a cloud computing system, server farm, and / or server cluster (or part thereof) that provides services to client devices or other devices or systems. In other examples, computing system 101 is one or more virtualized computer instances (e.g., virtual machines, containers, etc.) in a data center, cloud computing system, server farm, and / or server cluster, or is implemented by them.

[0014] User device 108 can be operated by a user. User device 108 may be implemented as any suitable client computing system, such as a mobile, non-mobile, wearable, and / or non-wearable computing device. User device 108 may be a smartphone, tablet computer, smartwatch, personal digital assistant, virtual assistant, gaming system, media player, e-book reader, television or television platform, laptop or notebook computer, desktop computer, camera, or any other type of wearable, non-wearable, mobile, or non-mobile computing device capable of performing one or more of the actions of the Disclosure.

[0015] Users such as architects and designers can interact with the BIM data generation system 130 to design a building. The BIM data generation system 130 can receive parameterized input data representing the building design, such as building parameters 124 or user input, and generate BIM data 132 using this input data. The BIM data 132 may include descriptive data of the building's structural features, such as building dimensions, number of floors, window locations and dimensions, and floor height; floor plans; structural information of the interior and exterior; 2D / 3D models of the building or their aspects; and civil, electrical, piping, lighting, or landscape data for the building. The BIM data generation system 130 may include computer-aided design (CAD) software to assist in the creation, modification, analysis, or optimization of the building design.

[0016] The building parameters 124 for a new building can be specifically generated to be processed by the BIM data generation system 130, which may be cross-platform, i.e., interchangeable between many different design systems and available to those systems. The BIM data generation system 130 may be a proprietary system having a specific set of input parameters. The building parameters 124 for a new building may be generated to conform to a set of input parameters for the BIM data generation system 130.

[0017] The BIM data generation system 130 may be implemented as any suitable computing system, such as one or more server computers, workstations, laptops, mainframes, appliances (dedicated equipment), cloud computing systems, smartphones, tablet computers, and / or other computing systems capable of performing the operations and / or functions described in accordance with one or more aspects of this disclosure. In some examples, the BIM data generation system 130 is a cloud computing system, server farm, and / or server cluster (or part thereof) that provides services to client devices or other devices or systems. In other examples, the BIM data generation system 130 is or is implemented by one or more virtualized computer instances (e.g., virtual machines, containers, etc.) in a data center, cloud computing system, server farm, and / or server cluster.

[0018] The computing system 101 and user device 108 may be the same single computing system or may be multiple different systems connected by a network. The BIM data generation system 130 and user device 108 may be the same single computing system or may be multiple different systems connected by a network. The computing system 101, BIM data generation system 130, and user device 108 may be the same single computing system or may be multiple different systems connected by a network. One or more networks connecting any of the systems of system 100 may be the Internet, any public or private communication network or other network, may include such networks, or may be part of such networks. For example, each network may be a cellular network, Wi-Fi®, ZigBee, Bluetooth® (or other personal area network (PAN)), near-field communication (NFC), ultrawideband, satellite, enterprise, service provider, and / or other types of networks that enable data transfer between network computing systems, servers, and computing devices. One or more of the client device, server device, and other devices may use any appropriate communication technology to send and receive data, commands, control signals, and / or other information over the network.

[0019] In the technology of this disclosure, a machine learning system 102 generates building parameters for a new building from a building image 114 which includes a depiction of an existing building. The user of the user device 108 may provide the building image 114 to the computing system 101 for processing. The building image 114 may be a photograph of an existing building. However, in some cases, the building image 114 may be a composite image, a simulated image, a sketch, a wireframe, an image of a 3D model, or other image that shows a building.

[0020] In inference mode, the machine learning system 102 may apply a trained machine learning model 103 to the building image 114 in order to predict the values ​​of building parameters 124 of the building indicated by the building image 114. The building parameters 124 may include descriptive data of the building's structural features, such as building dimensions, number of floors, window locations and dimensions, and floor height.

[0021] The machine learning model 103 may include one or more neural network models, each consisting of a neural network having one or more parameterized layers. Examples of neural networks include convolutional neural networks (CNNs), deep neural networks (DNNs), recurrent neural networks (RNNs), or combinations thereof. RNNs may be based on long- and short-term memory cells. Other types of machine learning models may be used.

[0022] In an example where the machine learning model 103 includes multiple layers, each layer may include a different set of artificial neurons. A layer may include an input layer, an output layer, and one or more hidden layers (also called intermediate layers). A layer may include fully connected layers, convolutional layers, pooling layers, and / or other types of layers. In a fully connected (i.e., "dense") layer, the output of each neuron in the previous layer forms the input of each neuron in the fully connected layer. In a convolutional layer, each neuron in the convolutional layer processes input from neurons associated with its receptive field. A pooling layer connects the outputs of neuron clusters in one layer into a single neuron in the next layer. Each input of each artificial neuron in each layer may be associated with a corresponding weight. Each artificial neuron may also have an activation function known in the art applied to it, such as a normalized linear unit (ReLU), hyperbolic tangent (TanH), or sigmoid.

[0023] As described above, in the example of FIG. 1, the machine learning model 103 includes a plurality of submodels (i.e., the image processing models 106 and one or more BIM data models 122) that are coupled to predict building parameters 124 from a building image 114. The image processing model 106 may include a trained CNN or other neural network that executes image processing operations on the building image 114. The image processing operations may include, for example, transformations of the building image 114 such as translation, rotation, tilt / distortion correction (deskewing), trapezoid correction (trapezoid processing), or scaling of the building shown in the building image 114. The image processing operations may also include detecting one or more facades (i.e., outer surfaces) of the building, and may further include identifying a main facade from the one or more detected facades. The "main facade" is usually the main front of the building including the front entrance, faces a major street or other thoroughfare, or is given as more important from an architectural point of view. The transformation operation may include generating an image of the facade of the building as viewed from a line perpendicular to the facade located at the center of the facade. In some cases, the transformation operation may include transforming the orientation of the image of the building to generate a front view of the building (an example of the generated image of the facade described above). The image processing model 106 may output the transformed building image 116 as the predicted image of the building, as shown in FIG. 1.

[0024] Also, the image processing operations may include identifying the building type of the building shown in the building image 114. The building type may include the category of the building, such as the architectural style or purpose of the building. Examples of architectural styles include modern, Gothic, Brutalism, Art Deco, neoclassicism, ranch style, Craftsman style, etc. Examples of the purpose of the building include lecture hall, apartment building, single-family house, stadium, place of worship, office building, etc. The building image may have one or more building types. For example, the building image may show a housing complex in a modern architectural style.

[0025] The image processing model 106 may output, as the building type 115, the prediction result regarding the building type of the building shown in the building image 114. In some examples, the machine learning system 102 may receive the building type of the building shown in the building image 114 in relation to the building image 114. The user may interact with the user device 108 to provide user input indicating the building type to be associated with the building image 114. The computing system 101 may obtain the building image 114 or the building type 115, for example, from a storage device or from the user device 108 via a network.

[0026] The machine learning system 102 applies one of the one or more BIM data models 122 to the building image to generate and output the building parameters 124 of the building shown by the building image. That is, the one or more BIM data models 122 generate values for the building parameters of the building, and those values are output by the machine learning system 102. The building parameters 124 may be output as BIM data and may be used to generate BIM data for parameterizing a new building. For example, the building parameters 124 may be input manually or automatically to the BIM data generation system 130, and the BIM data generation system 130 may generate BIM data 132 for a new building according to the building parameters 124. It is efficient because the parameterization of the BIM data 132 for the new building is performed by the building parameters 124 automatically generated from the building image. The BIM data 132 may be a 3D model, a wireframe model, a 2D model, a new building image, or another type of model. The BIM data 132 may, in some cases, be a partial model of a new building. In this way, the user or the computing system can use the predicted building parameters 124 to reproduce, for example, the exterior facade or the architectural style of the building in the building image 114 in the BIM data 132. Other types of BIM data are described elsewhere in this specification.

[0027] In various examples, the building image to which one of the BIM data models 122 is applied may be the received building image 114 or the converted building image 116.

[0028] In some examples, the BIM data model 122 is trained to predict building parameters 124 according to a specific building type or combination of building types. For example, BIM data model 122A may predict building parameters 124 for modern buildings, BIM data model 122B may predict building parameters 124 for Art Deco buildings, and BIM data model 122C may predict building parameters 124 for modern architectural housing complexes (i.e., multiple building types). In such examples, after applying the image processing model 106, the machine learning system 102 selects a BIM data model 122 corresponding to one or more building types of buildings shown in the building image 114. Again, the building types may be predicted by the image processing model 106 or received by the machine learning system 102 in relation to the building image 114. Returning to the example above, the machine learning system 102 may select BIM data model 122C to apply to a building image that is a modern architectural housing complex. The machine learning system 102 applies one or more selected BIM data models 122 to the building image in order to generate and output building parameters 124 of the building shown in the building image. In this case as well, the building image to which the BIM data model is applied may be the received building image 114 or the converted building image 116.

[0029] The machine learning system 102 may train the machine learning model 103 on the training data 104. The training data 104 may include images of buildings and corresponding known descriptive data for one or more building parameters of the building, such as building dimensions, number of floors, window locations and dimensions, floor height, and interior layout. The building images may be photographs, composite images, simulated images, sketches, wireframes of buildings, images of 3D models, or other images that depict buildings. The training data 104 may include data describing the materials of surface finishes and building elements such as walls, balconies, stairs, and fixtures, or exterior, interior, or both. The training data 104 may include data describing any structures that can be inferred from the building images, even if those structures are not included in the building images.

[0030] Machine learning can be supervised, semi-supervised, weakly supervised, or unsupervised. In semi-supervised learning, the dataset includes labeled and unlabeled data. An AI model is trained on a portion of the data (labeled data). The AI ​​model then predicts results for the unlabeled data. Only predictions with sufficient confidence are associated with the corresponding data as labels. This process is repeated until all data is labeled. Finally, supervised learning is performed using the obtained data. In weakly supervised learning, feedback on the prediction results is obtained on the application. Learning is then performed by repeatedly making predictions and receiving feedback. In unsupervised learning, representative vectors (prototypes) are obtained using clustering methods such as kmean. Building parameters are then set for each of these. In the technology disclosed herein, when a new building image 114 is input, the machine learning system 102 applies the machine learning model 103 to determine the closest prototype and obtain the corresponding building parameters.

[0031] The set of building parameters associated with building images in the training data 104 may be provided in a data structure such as a set of key-value pairs, a list, a table, or other associative data structure in which building parameters (e.g., "floor number") are associated with values ​​(e.g., "3"). Once the machine learning model 103 has been trained on the training data 104, the machine learning system 102 can apply the machine learning model 103 to predict the values ​​of building parameters 124 from building images 114.

[0032] In some examples, at least some of the building images in the training data 104 may be labeled with one or more building types corresponding to the buildings in those images. Thus, the image processing model 106 may be trained during inference mode to predict one or more building types of buildings indicated by the input building images 114. In such examples, the machine learning system 102 may train only one corresponding BIM data model 122 for its one or more building types. The machine learning system 102 can train one of the BIM data models 122 for a particular building type by training one of the BIM data models 122 on the training data 104 for that particular building type. Returning to the example above, for the building images in the training data 104 labeled "housing complex" and "modern architectural style," the machine learning system 102 may train a BIM data model 122C that is trained for a particular combination of building types. For the input building image 114, visual semantic embedding may be used to embed labels (image features) representing the building type into the feature space. In this case, the building parameters are predicted by applying the BIM data model 122 that is closest to the corresponding building type for each feature space.

[0033] In some examples, at least some of the building images in training data 104 are composite images. A program (such as one implemented by the BIM data generation system 130) may receive a set of inputs, each containing the values ​​of a set of building parameters. The program may then generate a composite image of the building according to the values ​​of this set of building parameters. Training data 104 may include the generated composite image associated with the values ​​of this set of building parameters. A user or another program may modify the input values, which are building parameters, within a range of dimensions, and input various sets of values ​​for a set of building parameters in order to generate multiple composite images of the building. Each of such generated composite images may become part of training data 104, along with the associated set of building parameter values ​​used to generate that composite image.

[0034] As a simple example, a building parameter may include two numbers, such as floors [1:3] and floor height [15 feet:18 feet]. In this case, the possible range for floors is [1:3], and the floor height is either 15 feet (approximately 4.57 m) or 18 feet (approximately 5.49 m). By manipulating the values ​​of possible combinations of (floors, floor height), the training data 104 can include six different composite images (i.e., (1, 15 feet), (2, 15 feet), (3, 15 feet), (1, 18 feet), (2, 18 feet), (3, 18 feet)) each associated with the values ​​of the building parameter used to generate the composite image. The building parameter may include one or more types of buildings, and their values ​​may become the labels of the corresponding generated composite images. With respect to the floor height [15 feet:18 feet] parameter in the example above, the two numbers do not necessarily represent possible values ​​for the floor height, but could also represent a possible range of floor heights, i.e., a minimum and maximum value. Therefore, the parameter "floor height [15 feet:18 feet]" can include any floor height between 15 feet and 18 feet.

[0035] Figure 2 shows an example of a computing system using the technology of this disclosure. Computing system 220 represents one or more computing devices configured for running machine learning system 224. Machine learning system 224 may be any example of a machine learning system in this disclosure, such as machine learning system 102 in Figure 1.

[0036] Memory 242 may store information for processing while the computing engine 222 is operating. In some examples, memory 242 includes temporary memory. "Temporary memory" means that the primary purpose of one or more of its storage devices is not long-term storage. Memory 242 is configured as volatile memory for short-term storage of information and therefore does not need to retain its stored contents when not in operation. Examples of volatile memory include random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), and other forms of volatile memory known in the art. Memory 242 also includes one or more computer-readable storage media in some examples. Memory 242 may be configured to store more information than volatile memory. Memory 242 is further configured as non-volatile memory space for long-term storage of information and may retain information after operation / shutdown cycles. Examples of non-volatile memory include magnetic hard disks, optical disks, floppy disks, flash memory, or EPROM or EEPROM. Memory 242 may store program instructions and / or data related to one or more modules described in accordance with one or more aspects of this disclosure. Memory 242 may store parameter weights of the machine learning model 103. In this example, the machine learning model 103 includes an image processing model 106 and a BIM data model 122.

[0037] The processing circuit 243 and memory 242 may provide an operating environment or platform for the computing engine 222. The computing engine 222 may be implemented as software, but in some examples it may include any combination of hardware, firmware, and software. The processing circuit 243 may execute instructions, and memory 242 may store instructions and / or data for one or more modules. The combination of the processing circuit 243 and memory 242 may retrieve, store, and / or execute instructions and / or data for one or more applications, modules, or software. The processing circuit 243 and memory 242 may also be operablely coupled to one or more other software and / or hardware components (including, but not limited to, one or more of the components shown in Figure 2).

[0038] The computing engine 222 may perform operations described using software, hardware, firmware, or a combination of hardware, software, and firmware that resides in or runs on the computing system 220. The computing engine 222 may run the machine learning system 224, or other programs and modules having multiple processors or multiple devices. The computing engine 222 may run the machine learning system 224 or other programs and modules as virtual machines or containers running on the underlying hardware. One or more such modules may run as one or more services of the operating system or computing platform. One or more such modules may run as one or more executable programs in the application layer of the computing platform.

[0039] One or more input devices 244 of the computing system 220 may generate, receive, or process inputs. Such inputs may include inputs from keyboards, pointing devices, voice response systems, video cameras, biometric detection / response systems, buttons, sensors, mobile devices, control pads, microphones, presence-sensing screens, networks, or any other type of device for detecting input from humans or machines.

[0040] One or more output devices 246 can generate, transmit, or process outputs. Examples of outputs include tactile, auditory, visual, and / or video outputs. Output devices 246 include displays, sound cards, video graphics adapter cards, speakers, presence-sensing screens, one or more USB interfaces, video and / or audio output interfaces, or any other type of device capable of generating tactile, audio, video, or other outputs. Output devices 246 may include display devices that can function as output devices using technologies including liquid crystal displays (LCDs), quantum dot displays, dot matrix displays, light-emitting diode (LED) displays, organic light-emitting diode (OLED) displays, cathode ray tube (CRT) displays, electronic inks, or any other type of display capable of generating monochrome, color, or tactile, auditory, and / or visual outputs. In some examples, the computing system 220 includes a presence-sensing display that functions as a user interface device acting as both one or more input devices 244 and one or more output devices 246.

[0041] One or more communication units 245 of the computing system 220 may communicate with external devices (or between individual computing devices of the computing system 220) by transmitting and / or receiving data, and in some examples may operate as both input and output devices. In some examples, the communication units 245 may communicate with other devices over a network. In other examples, the communication units 245 may transmit and / or receive radio signals over a radio network, such as a cellular radio network. Examples of communication units 245 include network interface cards (e.g., Ethernet® cards), optical transceivers, radio frequency transceivers, GPS receivers, or other types of devices capable of transmitting and / or receiving information. Other examples of communication units 245 include Bluetooth®, GPS, 3G, 4G, and Wi-Fi® radios found in mobile devices, and Universal Serial Bus (USB) controllers.

[0042] An input device 244 or communication unit 245 receives a building image 114. A machine learning model 103 may be used to generate a predicted output. A computing engine 222 runs a machine learning system 224 and applies it to the building image 114 to generate the predicted output in the form of building parameters 124. An output device 246 or communication unit 245 may output building parameters 124 to be used for designing a new building. In some examples, the input device 244 may include an image acquisition device, such as a camera, to generate the building image 114 as a photograph of the building. In such examples, the machine learning system 224, run by the computing system 220, receives the building image 114 from an integrated or separate image acquisition device.

[0043] Although the example in Figure 2 is described as being implemented using a neural network, the machine learning system 224 may, in addition to or instead of this, apply other types of machine learning to train one or more models. For example, the machine learning system 224 may apply one or more of the following supervised, unsupervised, semi-supervised, or reinforcement learning algorithms to train one or more models for prediction: nearest neighbor search, naive Bayes, decision trees, linear regression, support vector machines, neural networks, k-means algorithms, Q-learning, TD-learning, deep adversarial networks, or other supervised, unsupervised, semi-supervised, or reinforcement learning algorithms.

[0044] Figure 3 is a flowchart illustrating an example of a mode of operation for a machine learning system according to the technology described herein. While described in relation to the computing system 220 of Figure 2, which has a computing engine 222 that runs the machine learning system 224, mode of operation 300 may be performed by a computing system relating to other examples of machine learning systems described herein.

[0045] In operation mode 300, the computing engine 222 executes the machine learning system 102. The machine learning system 102 receives a building image 114 representing a building (302). The machine learning system 102 optionally (if it is the YES branch in 303) transforms the building image 114 to generate a front view image of the shown building and form a building image 116 (304). In the following description of operation mode 300, building images 114 and 116 are referred to as "images".

[0046] Next, the machine learning system 102 optionally processes the image to identify the building type of the building (306). Box 306 is indicated by a dotted line to show that this operation is optional. Next, the machine learning system 102 selects a BIM data model 122 and applies it to the image (114 or 116) to generate new building parameters for the new building, which are input to the BIM data generation system (308). In step 308, if a building type is identified, the machine learning system 102 may apply a BIM data model 122 that has been trained on building images of buildings that belong to the identified building type.

[0047] Next, the machine learning system outputs new building parameters for the new building (312).

[0048] The technologies described herein may be implemented, at least in part, by hardware, software, firmware, or any combination thereof. For example, various aspects of the technologies described herein may be implemented within one or more processors, including one or more microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), FPGAs, or any other equivalent integrated or discrete logic circuits or any combination of such components. The terms “processor” or “processing circuit” generally refer to any of the logic circuits described herein, any combination of such logic circuits with other logic circuits, or any other equivalent circuit. Furthermore, a control device including hardware may perform one or more of the technologies described herein.

[0049] Such hardware, software, and firmware may be implemented in the same device or in separate devices to support the various operations and functions described herein. In addition, any of the units, modules, or components described may be implemented as a single unit or as separate, interoperable logic circuits. The descriptions of the various features of a module or unit are intended to illustrate various functional aspects and do not necessarily imply that such a module or unit is implemented by separate hardware or software components. Rather, the functions associated with one or more modules or units may be performed by separate hardware or software components or integrated within the same or separate hardware or software components.

[0050] Furthermore, the technologies described herein may be embodied or encoded in a computer-readable medium that stores instructions, such as a computer-readable storage medium. Instructions embedded in or encoded in a computer-readable storage medium may, at execution time, cause a programmable processor or other processor to execute the method. Computer-readable storage media include random access memory (RAM), read-only memory (ROM), PROM, EPROM, EEPROM, flash memory, hard disks, CD-ROMs, floppy disks, cassettes, magnetic media, optical media, or other computer-readable media.

Claims

1. A machine learning system for predicting building parameters from building images, An input device configured to receive input including images of a building, A processing circuit and memory for executing the aforementioned machine learning system, It includes an output device, The aforementioned machine learning system is configured to apply a machine learning model, which has been trained using images of buildings and corresponding building parameters of those buildings, to images of buildings that have been received, in order to generate new building parameters for new buildings that are input into a BIM (building information modeling) data generation system. The output device is configured to output the new building parameters for the new building, The aforementioned machine learning model includes an image processing model and a BIM data model. The BIM data model includes multiple BIM data models for each building type. In order to apply the machine learning model to the received building image, the machine learning system: The image processing model is applied to the received image to generate an image of the front of the building. The image processing model is applied to the generated image of the building's facade to identify the building type. A machine learning system configured to apply a BIM data model corresponding to the identified building type from among the plurality of BIM data models to the generated front image to generate new building parameters for a building having the identified building type.

2. The machine learning system according to claim 1, wherein the learning of the image processing model is performed to correct the orientation of the received image of the building and generate an image of the front of the building.

3. A machine learning system for predicting building parameters for an image of a building, An input device configured to receive input including images of a building, A processing circuit and memory for executing the aforementioned machine learning system, It includes an output device, The aforementioned machine learning system is configured to apply a machine learning model, which has been trained using images of buildings and corresponding building parameters of those buildings, to images of buildings that have been received, in order to generate new building parameters for new buildings that are input into a BIM (building information modeling) data generation system. The output device is configured to output the new building parameters for the new building, The aforementioned machine learning model includes multiple BIM data models for each building type, In order to apply the machine learning model to the received building image, the machine learning system: The machine learning model is applied to the received image of the building to identify the building type. A machine learning system configured to apply a BIM data model corresponding to the identified building type from among the plurality of BIM data models to the received image to generate new building parameters for a building having the identified building type.

4. A machine learning system for predicting building parameters for an image of a building, An input device configured to receive input including images of a building, A processing circuit and memory for executing the aforementioned machine learning system, It includes an output device, The aforementioned machine learning system is configured to apply a machine learning model, which has been trained using images of buildings and corresponding building parameters of those buildings, to images of buildings that have been received, in order to generate new building parameters for new buildings that are input into a BIM (building information modeling) data generation system. The output device is configured to output the new building parameters for the new building, The machine learning system is further configured to receive the building type of the building, The aforementioned machine learning model includes multiple BIM data models for each building type, The machine learning system is configured to apply the machine learning model to a received image of a building, by applying the BIM data model corresponding to the building type of the building from among the plurality of BIM data models to the received image, thereby generating new building parameters for a building having the building type.

5. A machine learning system for predicting building parameters for an image of a building, An input device configured to receive input including images of a building, A processing circuit and memory for executing the aforementioned machine learning system, It includes an output device, The aforementioned machine learning system is configured to apply a machine learning model, which has been trained using images of buildings and corresponding building parameters of those buildings, to images of buildings that have been received, in order to generate new building parameters for new buildings that are input into a BIM (building information modeling) data generation system. The output device is configured to output the new building parameters for the new building, The machine learning system is configured to process the images of the building and the corresponding building parameters of the building in order to train the machine learning model. The aforementioned machine learning model includes multiple BIM data models for each building type, Each image of the building has a label that identifies the type of building, In order to process the image of the building and the corresponding building parameters of the building and train the machine learning model, the machine learning system shall For each image of the building, select the BIM data model from the plurality of BIM data models that corresponds to the label identifying the type of the building. A machine learning system configured to process the aforementioned images and corresponding building parameters of the aforementioned buildings in order to train the selected BIM data model.

6. The aforementioned image of the building is a composite image of the building. The machine learning system according to claim 5, wherein each of the composite images of the buildings is generated by inputting various values ​​for each of the building parameters into the program and causing the program to generate a composite image of the building for each combination of the values ​​of the building parameters.

7. Receiving images of the building, The machine learning system applies a machine learning model, trained using building images and corresponding building parameters, to the received building image to generate new building parameters for the new building, which are then input into the BIM (building information modeling) data generation system. The machine learning system also outputs new building parameters for the new building, The aforementioned machine learning model includes an image processing model and a BIM data model. The BIM data model includes multiple BIM data models for each building type. Applying the machine learning model to the received building image to generate the new building parameters is, The machine learning system applies the image processing model to the received image to generate an image of the front of the building. The machine learning system applies the image processing model to the generated image of the building's facade to identify the building type. A method comprising the machine learning system applying a BIM data model corresponding to the identified building type from among the plurality of BIM data models to the generated front image to generate new building parameters for a building having the identified building type.

8. The method according to claim 7, wherein the training of the image processing model is performed to correct the orientation of the received image of the building and generate an image of the front of the building.

9. Receiving an image of a building, The machine learning system applies a machine learning model, trained using building images and corresponding building parameters, to the received building image to generate new building parameters for the new building, which are then input into the BIM (building information modeling) data generation system. The machine learning system also outputs new building parameters for the new building, The aforementioned machine learning model includes multiple BIM data models for each building type, Applying the machine learning model to the received building image to generate new building parameters for the new building is, The machine learning model is applied to the received image of the building to identify the building type, A method comprising applying a BIM data model corresponding to the identified building type from among the plurality of BIM data models to the received image to generate new building parameters for a building having the identified building type.

10. Receiving an image of a building, The machine learning system applies a machine learning model, trained using building images and corresponding building parameters, to the received building image to generate new building parameters for the new building, which are then input into the BIM (building information modeling) data generation system. The machine learning system also outputs new building parameters for the new building, The machine learning system further comprises receiving the building type of the building, and the machine learning model includes a plurality of BIM data models for each building type. Applying the machine learning model to the received building image to generate new building parameters for the new building is, A method comprising applying a BIM data model corresponding to the building type of the building from among the plurality of BIM data models to the received image to generate new building parameters for a building having the building type of the building.

11. Receiving an image of a building, The machine learning system applies a machine learning model, trained using building images and corresponding building parameters, to the received building image to generate new building parameters for the new building, which are then input into the BIM (building information modeling) data generation system. The machine learning system also outputs new building parameters for the new building, The machine learning system further comprises processing the image of the building and the corresponding building parameters of the building to train the machine learning model, The aforementioned machine learning model includes multiple BIM data models for each building type, Each image of the building has a label that identifies the type of building, The above process is performed on each image of the building, The machine learning system selects from among the plurality of BIM data models the BIM data model corresponding to the label that identifies the type of building, A method comprising the machine learning system processing the image and corresponding building parameters of the building to train the selected BIM data model.

12. The method according to claim 11, wherein the image of the building is a composite image of the building, and each of the composite images of the building is generated by inputting various values ​​for each of the building parameters into a program and generating a composite image of the building for each combination of the values ​​of the building parameters.

13. The method according to claim 7, further comprising the BIM data generation system generating BIM data for the new building using the new building parameters for the new building.

14. The method according to claim 7, wherein the new building parameters include one or more of the following: building dimensions, number of floors, floor height, window location, and dimensions.

15. A non-temporary computer-readable medium containing machine-readable instructions for causing a processing circuit to perform an operation, wherein the operation is: Receiving images of the building, A machine learning model, trained using images of a building and corresponding building parameters, is applied to the received images of the building to generate new building parameters for a new building, which are then input into a BIM (building information modeling) data generation system. This includes outputting the new building parameters for the new building, The aforementioned machine learning model includes an image processing model and a BIM data model. The BIM data model includes multiple BIM data models for each building type. Applying the machine learning model to the received building image to generate the new building parameters is, The image processing model is applied to the received image to generate an image of the front of the building, The image processing model is applied to the generated image of the building's facade to identify the building type, A non-temporary computer-readable medium, comprising applying a BIM data model corresponding to the identified building type from among the plurality of BIM data models to the generated front image to generate new building parameters for a building having the identified building type.

16. The non-temporary computer-readable medium according to claim 15, wherein the learning of the image processing model is performed to correct the orientation of the received image of the building and generate an image of the front of the building.

17. A non-temporary computer-readable medium including a machine-readable instruction for causing a processing circuit to perform an operation, wherein the operation is: Receiving images of the building, A machine learning model, trained using images of a building and corresponding building parameters, is applied to the received images of the building to generate new building parameters for a new building, which are then input into a BIM (building information modeling) data generation system. This includes outputting the new building parameters for the new building, The aforementioned machine learning model includes multiple BIM data models for each building type, Applying the machine learning model to the received building image to generate new building parameters for the new building is, The machine learning model is applied to the received image of the building to identify the building type, A non-temporary computer-readable medium, comprising applying a BIM data model corresponding to the identified building type from among the plurality of BIM data models to the received image to generate new building parameters for a building having the identified building type.

18. A non-temporary computer-readable medium including a machine-readable instruction for causing a processing circuit to perform an operation, wherein the operation is: Receiving images of the building, A machine learning model, trained using images of a building and corresponding building parameters, is applied to the received images of the building to generate new building parameters for a new building, which are then input into a BIM (building information modeling) data generation system. This includes outputting the new building parameters for the new building, The operation further includes receiving the building type of the building, and the machine learning model includes a plurality of BIM data models for each building type. Applying the machine learning model to the received building image to generate new building parameters for the new building is, A non-temporary computer-readable medium, comprising applying a BIM data model corresponding to the building type of the building from among the plurality of BIM data models to the received image to generate new building parameters for a building having the building type of the building.

19. A non-temporary computer-readable medium including a machine-readable instruction for causing a processing circuit to perform an operation, wherein the operation is: Receiving images of the building, A machine learning model, trained using images of a building and corresponding building parameters, is applied to the received images of the building to generate new building parameters for a new building, which are then input into a BIM (building information modeling) data generation system. This includes outputting the new building parameters for the new building, The operation further includes processing the image of the building and the corresponding building parameters of the building to train the machine learning model, The aforementioned machine learning model includes multiple BIM data models for each building type, Each image of the building has a label that identifies the type of building, The above process is performed on each image of the building, Selecting a BIM data model from among the plurality of BIM data models that corresponds to the label that identifies the type of the building, A non-temporary computer-readable medium, which includes processing the aforementioned image and corresponding building parameters of the aforementioned building to train the selected BIM data model.

20. The non-temporary computer-readable medium according to claim 19, wherein the image of the building is a composite image of the building, and each of the composite images of the building is generated by inputting various values ​​for each of the building parameters into a program and generating a composite image of the building for each combination of the values ​​of the building parameters.

Images