Learning device, learning method, and program

Abstract

To generate a machine learning model that accurately detects a road area from a far area, out of areas included in an image.SOLUTION: A learning device that trains a machine learning model that receives an image including a plurality of pixels as an input and outputs a determination value indicating whether each of the pixels represents a road, the learning device including: a setting unit configured to set a weight for an error between the determination value output by the machine learning model and learning data indicating whether each of the pixels represents a road; and a learning unit configured to train the machine learning model so as to reduce a value of a loss function calculated on the basis of the error with the weight set thereto. The setting unit is configured to increase the weight in one or more predetermined directions from a lower end center of the image.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a learning device, a learning method, and a program. [Background technology]

[0002] Conventionally, there are known techniques for detecting road areas included in an image in order to utilize the technique for driving assistance or autonomous driving of a vehicle. For example, Patent Document 1 discloses a technique for detecting a road area by scanning an input image in the x-direction, extracting edges, and detecting white lines based on color differences near the edges. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-038101 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the technology described in Patent Document 1, the number of pixels decreases as the distance increases in the area included in the image, which may result in a decrease in scanning accuracy. Thus, in the conventional technology, there are cases where road areas cannot be accurately detected from distant areas included in the image.

[0005] The present invention has been made in consideration of these circumstances, and one of its objectives is to provide a learning device, a learning method, and a program capable of generating a machine learning model that accurately detects road areas from distant areas contained in an image. [Means for solving the problem]

[0006] The learning device, the learning method, and the program according to the present invention employ the following configuration. (1): A learning device according to one embodiment of the present invention is a learning device that learns a machine learning model that receives an image including a plurality of pixels as input and outputs a discriminant value indicating whether each of the pixels represents a road or not, and includes a setting unit that sets a weight for an error between the discriminant value output by the machine learning model and learning data indicating whether each of the pixels represents a road or not, and a learning unit that trains the machine learning model so as to reduce a value of a loss function calculated based on the error for which the weight is set, and the setting unit increases the weight from the center of the bottom end of the image toward one or more specified directions.

[0007] (2) In the above aspect (1), the setting unit increases the weight from the center of the bottom edge of the image toward the top edge and the left and right edges as the one or more predetermined directions.

[0008] (3): A learning method according to another aspect of the present invention is a learning method for training a machine learning model that uses an image including a plurality of pixels as input and outputs a discriminant value indicating whether each of the pixels represents a road or not, in which a computer sets a weight for an error between the discriminant value output by the machine learning model and training data indicating whether each of the pixels represents a road or not, and trains the machine learning model so as to reduce the value of a loss function calculated based on the error for which the weight is set, and the setting increases the weight from the center of the bottom edge of the image toward one or more specified directions.

[0009] (4): A program according to another aspect of the present invention is a program for training a machine learning model that receives an image including a plurality of pixels as input and outputs a discriminant value indicating whether each of the pixels represents a road or not, the program instructs a computer to set a weight for the error between the discriminant value output by the machine learning model and training data indicating whether each of the pixels represents a road or not, and trains the machine learning model to reduce the value of a loss function calculated based on the error for which the weight has been set, the setting increasing the weight from the center of the bottom edge of the image toward one or more specified directions. Effect of the Invention

[0010] According to (1) to (4), a machine learning model can be generated that can accurately detect road areas from distant areas among the areas included in an image. [Brief description of the drawings]

[0011] [Figure 1] 1 is a diagram illustrating a configuration of a learning device 100 according to the present embodiment. [Diagram 2] FIG. 2 is a diagram showing an example of the configuration of learning data 130A. [Diagram 3] 10 is a diagram for explaining a method for setting weights by the setting unit 110. FIG. [Figure 4] 4 is a flowchart showing an example of the flow of processing executed by the learning device 100 according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, embodiments of a learning device, a learning method, and a program according to the present invention will be described with reference to the drawings.

[0013] [composition] FIG. 1 is a diagram showing a configuration of a learning device 100 according to this embodiment. The learning device 100 is an information processing device that learns a machine learning model that receives an image including a plurality of pixels as an input and outputs a discriminant value indicating whether each pixel represents a road. The learning device 100 includes, for example, a setting unit 110, a learning unit 120, and a storage unit 130. The setting unit 110 and the learning unit 120 are each realized by, for example, a hardware processor such as a CPU (Central Processing Unit) executing a program (software). In addition, some or all of these components may be realized by hardware (including circuitry) such as an LSI (Large Scale Integration), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a GPU (Graphics Processing Unit), or may be realized by cooperation between software and hardware. The program may be stored in advance in a storage device (a storage device equipped with a non-transient storage medium) such as the HDD or flash memory of the learning device 100, or may be stored in a removable storage medium such as a DVD or CD-ROM, and installed in the HDD or flash memory of the learning device 100 by mounting the storage medium (non-transient storage medium) in a drive device. The storage unit 130 stores, for example, learning data 130A and machine learning model 130B. The storage unit 150 is realized by, for example, a RAM, a flash memory, an SD card, or the like.

[0014] FIG. 2 is a diagram showing an example of the configuration of the learning data 130A. The learning data 130A is, for example, a label value indicating whether or not a pixel represents a road is associated with a pixel PX (pixel group) of one or more images IM. For example, in the case of FIG. 2, since the area between the road dividing lines LRL and RRL represents a road area, the pixel PX1 is associated with a label value 0 indicating that the pixel does not represent a road, while the pixels PX2 to PX11 are associated with a label value 1 indicating that the pixel represents a road. The learning data 130A is generated, for example, by an administrator or operator of the learning device 100 specifying each pixel and a label value of the image IM in advance on his / her own terminal, and is stored in the storage unit 150. Alternatively, the learning device 100 may download the learning data 130A stored in an external server to the storage unit 130 via a network at the timing of execution of learning by the learning unit 120. The label value is an example of a "discrimination value" in the claims.

[0015] The machine learning model 130B is a machine learning model that receives an image IM as input and outputs a label value indicating whether or not a pixel of the image IM represents a road. The machine learning model 130B is, for example, a convolutional neural network (CNN), which extracts features of a pixel (a group of pixels) of the input image IM to output a label value indicating whether or not the pixel represents a road. The image IM input to the machine learning model 130B may be a raw image captured by a camera mounted on a vehicle, or may be a planar projection image or a cylindrical projection image of a road obtained by converting the raw image.

[0016] Here, in the conventional technology, the machine learning model is generally trained so as to reduce the value of a loss function calculated based on the error between the label value output by the machine learning model and the training data 130A indicating whether each pixel represents a road, and a final trained model is generated. Here, the loss function is defined as, for example, the absolute sum of the errors for each pixel. On the other hand, the further away from the point where the image IM is photographed, the fewer the number of pixels that represent an area of ​​the same area. For example, as shown in the right part of FIG. 2, although the areas A1 and A2 have the same area on a real road, the area A1 that is farther from the photographing point is represented by a single pixel PX2, while the area A2 that is closer to the photographing point is represented by multiple pixels PX3 to PX11. As a result, the trained model generated by the conventional technology has a high output accuracy for the area of ​​the image IM that is closer to the photographing point, but tends to have a low output accuracy for the area that is farther from the photographing point.

[0017] In light of these circumstances, the setting unit 110 sets (for example, multiplies) a weight for the error between the label value output by the machine learning model 130B and the learning data 130A indicating whether each pixel represents a road. More specifically, the setting unit 110 increases the weight value set for the error from the center of the bottom edge of the image IM toward one or more predetermined directions. That is, the setting unit 110 increases the value of the error for an area farther from the shooting point by using the weight compared to the error for an area closer to the shooting point, thereby preventing a decrease in the accuracy of machine learning for distant areas.

[0018] FIG. 3 is a diagram for explaining a method of setting weights by the setting unit 110. The setting unit 110 increases the weight from the center of the lower end of the image IM toward the upper end and the left and right ends as one or more predetermined directions. For example, as shown in the pattern of FIG. 3(a), the setting unit 110 may increase the weight value set for the error from the center of the lower end of the image IM toward the upper end, as shown in the pattern of FIG. 3(b), the setting unit 110 may increase the weight value set for the error from the center of the lower end of the image IM toward the right end, as shown in the pattern of FIG. 3(c), the setting unit 110 may increase the weight value set for the error from the center of the lower end of the image IM toward the left end. These patterns may be applied simultaneously (for example, when a pixel is located in the upper left direction of the image IM, both the weight of the pattern of FIG. 3(a) and the weight of the pattern of FIG. 3(c) may be set), or only a specific pattern may be selected and applied from these patterns. In this embodiment, the setting unit 110 increases the weight value set for the error at least from the center of the bottom edge of the image IM toward the top edge.

[0019] After the setting unit 110 sets weights for errors between the label values ​​output by the machine learning model 130B and the learning data 130A indicating whether each pixel of the image IM represents a road or not, the learning unit 120 trains the machine learning model 130B so as to reduce the value of a loss function (e.g., absolute sum of errors) based on the errors for which the weights are set. In this case, any learning method such as back propagation may be used for the learning. The learning unit 120 acquires the machine learning model 130B for which learning has been completed as a trained model.

[0020] The acquired trained model may be used by a driving assistance device or a vehicle control device mounted on the vehicle. For example, the driving assistance device may input a forward image of the vehicle captured by a camera into the trained model, and display the road area output by the trained model on the HMI. Also, for example, the vehicle control device may input a forward image of the vehicle captured by a camera into the trained model, and control the vehicle to travel within the road area output by the trained model.

[0021] Thus, in this embodiment, the setting unit 110 increases the weighting value set for the error from the center of the bottom edge of the image IM toward one or more predetermined directions, and the learning unit 120 learns the machine learning model 130B so as to reduce the value of the error function based on the error for which the weighting is set. This makes it possible to generate a machine learning model that accurately detects road areas from distant areas among areas included in an image.

[0022] [Processing flow] Next, the flow of processing executed by the learning device 100 according to this embodiment will be described with reference to Fig. 4. Fig. 4 is a flowchart showing an example of the flow of processing executed by the learning device 100 according to this embodiment.

[0023] First, the learning unit 120 acquires an image from the learning data 130A stored in the storage unit 130 (step S100). Next, the learning unit 120 calculates, for each pixel of the acquired image, an error between the correct label of the learning data 130A and the output of the machine learning model 130B (step S102).

[0024] Next, the setting unit 110 sets a weight for the error calculated by the learning unit 120 according to the distance from the center of the bottom edge of the image (step S104). Next, the learning unit 120 calculates a loss function based on the error for which the weight has been set (step S106). Next, the learning unit 120 trains the machine learning model 130B so as to reduce the value of the calculated loss function (step S108). This ends the process according to this flowchart.

[0025] As described above, in this embodiment, a weight is set for the error between the discrimination value output by the machine learning model that receives an image including a plurality of pixels as input and outputs a discrimination value indicating whether each pixel represents a road or not and the learning data indicating whether each pixel represents a road or not, and the machine learning model is trained to reduce the value of the loss function calculated based on the error for which the weight is set, and the setting is such that the weight is increased from the center of the bottom edge of the image in one or more predetermined directions. This makes it possible to generate a machine learning model that can accurately detect road areas from distant areas among areas included in an image.

[0026] The above-described embodiment can be expressed as follows. A storage device storing a program; a hardware processor; The hardware processor executes the program, a machine learning model that receives an image including a plurality of pixels as input and outputs a discrimination value indicating whether each of the pixels represents a road or not, and sets a weight for an error between the discrimination value output by the machine learning model and learning data indicating whether each of the pixels represents a road or not; training the machine learning model so as to reduce a value of a loss function calculated based on the error to which the weight is set; The setting increases the weight from the center of the bottom edge of the image toward one or more predetermined directions. The learning device is configured as follows.

[0027] The above describes the form for carrying out the present invention using an embodiment, but the present invention is not limited to such an embodiment, and various modifications and substitutions can be made within the scope that does not deviate from the gist of the present invention. [Explanation of symbols]

[0028] 100 Learning Device 110 Setting section 120 Learning Department 130 Storage section 130A Training Data 130B Machine Learning Model

Claims

1. A learning device that learns a machine learning model that receives an image including a plurality of pixels as input and outputs a discriminant value indicating whether each of the pixels represents a road, comprising: a setting unit that sets a weight for an error between a discrimination value output by the machine learning model and learning data indicating whether each of the pixels represents a road; A learning unit that learns the machine learning model so as to reduce a value of a loss function calculated based on the error to which the weight is set, the setting unit increases the weight from a center of a lower end of the image toward one or more predetermined directions. Learning device.

2. the setting unit increases the weight from a center of a lower end of the image toward an upper end and left and right ends of the image as the one or more predetermined directions. The learning device according to claim 1 .

3. the setting unit increases the weight from a center of a lower end of the image toward an upper end of the image as the one or more predetermined directions. The learning device according to claim 1 .

4. A learning method for learning a machine learning model that receives an image including a plurality of pixels as input and outputs a discriminant value indicating whether each of the pixels represents a road, the method comprising the steps of: setting a weight for an error between the discriminant value output by the machine learning model and learning data indicating whether each of the pixels represents a road; training the machine learning model so as to reduce a value of a loss function calculated based on the error to which the weight is set; The setting increases the weight from the center of the bottom edge of the image toward one or more predetermined directions. How to learn.

5. A program for training a machine learning model that receives an image including a plurality of pixels as input and outputs a discriminant value indicating whether each of the pixels represents a road, the program comprising: setting a weight for an error between a discriminant value output by the machine learning model and learning data indicating whether each of the pixels represents a road; training the machine learning model so as to reduce a value of a loss function calculated based on the error to which the weight is set; The setting increases the weight from the center of the bottom edge of the image toward one or more predetermined directions. program.

Images