DATA GENERATION DEVICE, DATA GENERATION METHOD, AND PROGRAM

Abstract

To make it easier to prepare learning data for generating an image recognition model.SOLUTION: An image obtainment section 110 obtains a plurality of images. The plurality of images includes an object to be estimated. An image carve out section 120 carves out an object region including the object from each of the plurality of images obtained by the image obtainment section 110. An importance level generation section 130 processes the object region carved out by the image carve out section 120, for generating importance level information. The importance level information indicates an importance level of the object region when generating an object estimation model, and is generated for each object region, thus for each image obtained by the image obtainment section 110. A learning data generation section 140 stores a plurality of the object regions curved out by the image carve out section 120 and a plurality of pieces of the importance level information generated by the importance level generation section 130 into a learning data storage section 150 as at least a part of learning data.SELECTED DRAWING: Figure 2

Description

[Technical field]

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

[0002] In recent years, machine learning has been used in various fields. In order to generate a model by machine learning, it is necessary to prepare learning data. Patent Document 1 describes that when applying machine learning to the control of a robot used in a factory, learning data is prepared using the following method. Specifically, first, object information of the target is associated with a marker for detecting position and orientation. In addition, a learning data set generation tool is prepared. This learning data set generation tool is composed of a base part that serves as a guide for the placement position of the target, and a marker fixed above the base part. Then, in a state in which the target is placed using the base part as a guide, a group of multi-viewpoint images of the entire object including the marker is acquired. Then, a bounding box of the target is set for the acquired image group, and orientation information of the target, center of gravity position information, object information, and information on the bounding box estimated from the captured image are associated with the captured image, thereby generating a learning data set for performing object recognition and position and orientation estimation of the target. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2019-29021 A Summary of the Invention [Problem to be solved by the invention]

[0004] In order to improve the accuracy of a model generated by machine learning, it is necessary to prepare a large amount of training data. One example of the objective of the present invention is to facilitate the preparation of training data that is used when generating a model for image recognition and has a high learning effect. [Means for solving the problem]

[0005] According to the present invention, there is provided a data generation device for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: image capture means for capturing a plurality of images including the object; an image extraction means for extracting an object region including an object from each of the plurality of images; an importance generating means for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating means for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; A data generating device is provided, comprising:

[0006] According to the present invention, there is provided a data generation method for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising the steps of: The computer an image capture process for capturing a plurality of images including the object; an image extraction process for extracting an object region including an object from each of the plurality of images; an importance information generation process for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generation process for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; A data generation method is provided that performs the above.

[0007] According to the present invention, there is provided a program for causing a computer to function as a data generation device for generating learning data for generating an object estimation model for estimating a type of an object included in an image, the program comprising: The computer includes: an image capture function for capturing a plurality of images including the object; an image cropping function for cropping an object region including an object from each of the plurality of images; an importance information generating function that generates importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating function for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; A program will be provided to help students develop these skills. Effect of the Invention

[0008] According to the present invention, it is easy to prepare learning data that is used when generating a model for image recognition and has a high learning effect. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating a usage environment of a data generating device according to a first embodiment. [Diagram 2] FIG. 2 illustrates an example of a functional configuration of a data generating device. [Diagram 3] FIG. 2 is a diagram for explaining a first example of a process performed by an image cropping unit. [Figure 4] 13 is a diagram for explaining a second example of the process performed by the image cropping unit. FIG. [Diagram 5] FIG. 2 illustrates an example of a hardware configuration of a data generating device. [Figure 6] 13 is a flowchart illustrating an example of a process performed by a data generating device. [Figure 7] FIG. 11 is a diagram illustrating a functional configuration of a data generating device according to a second embodiment. [Figure 8] FIG. 11 is a diagram illustrating a functional configuration of a data generating device according to a third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In all the drawings, the same components are given the same reference numerals and the description will be omitted as appropriate.

[0011] (First embodiment) FIG. 1 is a diagram for explaining a usage environment of a data generating device 10 according to this embodiment. The data generating device 10 is used together with an object recognition device 20. The object recognition device 20 processes an image using an object estimation model to estimate (identify) the type of an object contained in the image. The data generating device 10 generates learning data for generating the object estimation model. In this embodiment, the data generating device 10 also generates the object estimation model. Then, the data generating device 10 transmits the generated object estimation model to the object recognition device 20. The object recognition device 20 may transmit the model to one object recognition device 20, or may transmit the object estimation model to multiple object recognition devices 20.

[0012] The object recognition device 20 is, for example, a product registration device used at checkout in a store. In this case, the object to be estimated is a product. The object recognition device 20 generates an image to be processed by capturing an image of the product to be registered. The object recognition device 20 processes this image to identify the product and register it as a product to be checked out. The object recognition device 20 may be operated by a store clerk or a customer. The object recognition device 20 may or may not have a checkout function. In the latter case, the object recognition device 20 is used together with a checkout device.

[0013] However, the object recognition device 20 is not limited to a merchandise registration device.

[0014] 2 is a diagram showing an example of a functional configuration of the data generating device 10. The data generating device 10 includes an image acquiring unit 110, an image cropping unit 120, an importance generating unit 130, and a learning data generating unit 140.

[0015] The image acquisition unit 110 acquires a plurality of images. The plurality of images includes an object to be estimated. In the example shown in the figure, at least some (or all) of the plurality of images are a plurality of frame images included in one video. This video is generated, for example, by capturing an object while changing the relative orientation of the object with respect to the imaging device. In this way, each of the plurality of frame images included in one video is a still image of the object captured from a different direction. Therefore, the object recognition model generated by the data generation device 10 can accurately estimate the type of the object even if the object is not facing in a specific direction.

[0016] When generating a video, the object may be photographed while being held in a person's hand. In this case, the person changes the orientation of the object during shooting. On the other hand, when generating a video, the object may be stationary. In this case, the imaging device moves during shooting, changing the orientation of the object relative to the imaging device.

[0017] The image cropping unit 120 crops out an object region including an object from each of the multiple images acquired by the image acquisition unit 110. An example of the shape of the object region is a rectangle. In this embodiment, the type of object to be recognized is known in advance. Therefore, the image cropping unit 120 may specify and crop out the object region using the feature amount of this object. Other examples of the processing performed by the image cropping unit 120 will be described later with reference to other figures.

[0018] The importance generation unit 130 generates importance information by processing the object region cut out by the image cutout unit 120. The importance information indicates the importance of the object region when generating an object estimation model, and is generated for each object region, i.e., for each image acquired by the image acquisition unit 110. The importance information includes, for example, a score indicating the importance.

[0019] As an example, the importance generating unit 130 generates importance information using the sharpness of an object in the object region. The sharpness decreases as the shaking or out-of-focus (blur) during shooting increases. Specifically, when the sharpness of an object is low, the importance generating unit 130 decreases the importance of the object region. The sharpness can be calculated using, for example, an autofocus technique used in the shooting device. Specifically, when the object is in focus, the sharpness is high. Furthermore, the image cropping unit 120 may use the gradient of color information of pixels (for example, a numerical value indicating the strength of each of RGB) at the edge of the object as a value indicating the sharpness. Then, the importance generating unit 130 determines that the sharpness is low as the gradient becomes smaller.

[0020] As another example, the importance generation unit 130 generates importance information using information about the size of an object in an object region. For example, the importance generation unit 130 reduces the importance of an object region as the object in the object region becomes smaller. Also, when a part of an object is hidden, the importance generation unit 130 reduces the importance of the object region.

[0021] As another example, when an object is held by a person's hand, the importance generating unit 130 generates importance information using information about the size of the hand in the object region. This is because when the proportion of the hand occupies a large proportion of the object region, the proportion of the object occupies a small proportion of the object region. Specifically, the importance generating unit 130 lowers the importance of the object region as the size of the hand in the object region increases.

[0022] The learning data generation unit 140 stores the multiple object regions cut out by the image cutout unit 120 and the multiple pieces of importance information generated by the importance generation unit 130 in the learning data storage unit 150 as at least a part of the learning data. At this time, each of the multiple object regions is linked to the importance information of the object region. In this learning data, the object region is used as at least a part of the explanatory variables. When learning an object region with low importance, the learning data generation unit 140 reduces the influence of the object region on the object recognition model (for example, the contribution to loss). The objective variable of this learning data is the type of the object (for example, the product name).

[0023] The video storage unit 112 stores videos for each of a plurality of objects. The image cropping unit 120, the importance generating unit 130, and the learning data generating unit 140 perform the above-described processing for each object.

[0024] In the example shown in the figure, the data generating device 10 further includes a model generating unit 160, a model storage unit 170, and a transmission unit 180. The model generating unit 160 generates an object estimation model using the learning data stored in the learning data storage unit 150. Here, the model generating unit 160 generates one object estimation model using the learning data generated for each type of object. When an image including an object is input, this object estimation model outputs an estimation result of the type of object. The model storage unit 170 stores the object estimation model generated by the model generating unit 160. The transmission unit 180 transmits the object estimation model stored in the model storage unit 170 to the object recognition device 20.

[0025] In the example shown in the figure, the video storage unit 112, the learning data storage unit 150, and the model storage unit 170 are part of the data generating device 10. However, at least one of these storage units may be provided outside the data generating device 10.

[0026] Furthermore, the model generating unit 160, the model storage unit 170, and the transmitting unit 180 may be provided in a device separate from the data generating device .

[0027] 3 is a diagram for explaining an example of processing performed by the image cropping unit 120. In the example shown in this figure, an object is held in a hand. The image cropping unit 120 first identifies the position of the person's hand and identifies the object region using this hand position. In this case, the image cropping unit 120 identifies the hand position using, for example, a feature amount of the person's hand, and detects multiple positions (object position candidates) where an object may exist using, for example, an object detection model. Then, the position closest to the hand position among the multiple object position candidates is recognized as the object region.

[0028] 4 is a diagram for explaining another example of the process performed by the image cropping unit 120. In the example shown in this figure, a code such as a barcode or a two-dimensional code is attached to the object. The image cropping unit 120 then identifies the position of this code and identifies the object region using the position of this code. In this case as well, the image cropping unit 120 identifies the position of the code using, for example, a feature amount of the code, and identifies the object including this position using, for example, edge detection processing. Then, the region including this object is recognized as the object region.

[0029] 5 is a diagram showing an example of a hardware configuration of the data generating device 10. The data generating device 10 includes a bus 1010, a processor 1020, a memory 1030, a storage device 1040, an input / output interface 1050, and a network interface 1060.

[0030] The bus 1010 is a data transmission path for transmitting and receiving data among the processor 1020, memory 1030, storage device 1040, input / output interface 1050, and network interface 1060. However, the method of connecting the processor 1020 and the like to each other is not limited to bus connection.

[0031] The processor 1020 is implemented by a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or the like.

[0032] The memory 1030 is a main storage device realized by a RAM (Random Access Memory) or the like.

[0033] The storage device 1040 is an auxiliary storage device realized by a hard disk drive (HDD), a solid state drive (SSD), a memory card, a read only memory (ROM), or the like. The storage device 1040 stores program modules that realize each function of the data generating device 10 (for example, the image acquiring unit 110, the image cropping unit 120, the importance generating unit 130, the learning data generating unit 140, the model generating unit 160, and the transmitting unit 180). The processor 1020 loads each of these program modules onto the memory 1030 and executes them, thereby realizing each function corresponding to the program module. The storage device 1040 also functions as the video storage unit 112, the learning data storage unit 150, and the model storage unit 170.

[0034] The input / output interface 1050 is an interface for connecting the data generating device 10 to various input / output devices.

[0035] The network interface 1060 is an interface for connecting the data generating device 10 to a network. This network is, for example, a LAN (Local Area Network) or a WAN (Wide Area Network). The method for connecting the network interface 1060 to the network may be a wireless connection or a wired connection. The data generating device 10 may communicate with the object recognition device 20 via the network interface 1060.

[0036] 6 is a flowchart showing an example of processing performed by the data generating device 10. The data generating device 10 performs the processing shown in this figure for each type of object.

[0037] First, the image acquisition unit 110 of the data generating device 10 acquires a moving image of an object to be processed (step S110). Then, the data generating device 10 performs the processes shown in steps S120 to S150 on all of the frame images constituting the acquired moving image (step S160).

[0038] First, the image cropping unit 120 selects a frame image to be processed (step S120). Then, the image cropping unit 120 crops out an object area from the selected frame image (step S130). Details of this process are as described with reference to Figs. 2 to 4. Although not shown, if an object is not captured in the selected frame image, the data generating device 10 proceeds to step S160.

[0039] Next, the importance generation unit 130 processes the object region to generate importance information (step S140). Details of this process are as described with reference to Fig. 2. Then, the learning data generation unit 140 updates the learning data by storing the object region and the importance information in the learning data storage unit 150 (step S150).

[0040] As described above, according to this embodiment, the learning data generated by the data generating device 10 includes an image (object region) including an object, as well as importance data indicating the importance of the object region. Therefore, even if the learning data generated by 10 includes an image that is not suitable as learning data (e.g., a blurred image), the influence of the image on the model is small. Therefore, even if an object recognition model is generated using learning data generated by the data generating device 10, the accuracy of the object recognition model is high. Therefore, by using the data generating device 10, learning data can be easily generated.

[0041] Second embodiment Fig. 7 is a diagram showing the functional configuration of the data generating device 10 according to this embodiment, and corresponds to Fig. 2 of the first embodiment. The data generating device 10 shown in this figure is similar to the data generating device 10 according to the first embodiment, except that the image cropping unit 120 has a tracking unit 122.

[0042] After identifying the position of an object in a certain frame image (hereinafter referred to as the first frame image), the tracking unit 122 uses the position of the object in the first frame image when identifying the position of the object in a subsequent frame image (hereinafter referred to as the second frame image). That is, the tracking unit 122 has a function of tracking an object in a video. As an example, the tracking unit 122 searches for an object, focusing on an area near the position of the object in the first frame image. For example, the second frame image is the frame image next to the first frame image, but is not limited to this.

[0043] In this embodiment, similarly to the first embodiment, learning data can be easily generated by using the data generating device 10. In addition, since the image cutting unit 120 has the tracking unit 122, an object region can be easily cut out from an image.

[0044] Third embodiment Fig. 8 is a diagram showing the functional configuration of the data generating device 10 according to this embodiment, and corresponds to Fig. 2 of the first embodiment. The data generating device 10 shown in this figure is similar to the data generating device 10 according to the first embodiment, except for the following points.

[0045] The importance generation unit 130 generates importance information using a model for calculating importance (hereinafter, importance estimation model). The importance estimation model outputs importance information when an object region is input. The importance estimation model is stored in, for example, an estimation model storage unit 132. The estimation model storage unit 132 may be a part of 10 or may be located outside the data generating device 10.

[0046] The importance estimation model is generated, for example, as follows. First, learning data for the importance estimation model is prepared. This learning data has an image corresponding to an object region as an explanatory variable and importance information as a target variable. When generating the learning data, the importance information is input, for example, by a person. Next, machine learning is performed using this learning data. As a result, the importance estimation model is generated.

[0047] In this embodiment, the image cropping unit 120 may also include the tracking unit 122 shown in the second embodiment.

[0048] In this embodiment, similarly to the first embodiment, training data can be easily generated by using the data generating device 10.

[0049] Although the embodiments of the present invention have been described above with reference to the drawings, these are merely examples of the present invention, and various configurations other than those described above can also be adopted.

[0050] In addition, in the flowcharts used in the above description, multiple steps (processes) are described in order, but the order of execution of the steps performed in each embodiment is not limited to the order described. In each embodiment, the order of the steps shown in the figures can be changed to the extent that the content is not affected. In addition, the above-mentioned embodiments can be combined to the extent that the content is not contradictory.

[0051] A part or all of the above-described embodiments can be described as, but are not limited to, the following supplementary notes. 1. A data generation device for generating learning data for generating an object estimation model for estimating a type of an object included in an image, comprising: image capture means for capturing a plurality of images including the object; an image extraction means for extracting an object region including an object from each of the plurality of images; an importance generating means for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating means for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; A data generating device comprising: 2. In the data generating device according to the above item 1, A data generation device in which the image acquisition means acquires a plurality of frame images included in a single video as at least a portion of the plurality of images. 3. In the data generating device according to the above item 2, The image extraction means is a data generating device that uses the position of the object area in a first frame image to extract the object area from a second frame image that is generated after the first frame image. 4. In the data generating device according to any one of claims 1 to 3, A data generating device comprising: a model generating means for generating the object estimation model using the training data. 5. In the data generating device according to any one of 1 to 4 above, The importance generating means is a data generating device that generates the importance information by using the sharpness of the object in the object region. 6. In the data generating device according to any one of claims 1 to 5, The importance generating means is a data generating device that generates the importance information using information regarding the size of the object in the object region. 7. In the data generating device according to any one of claims 1 to 6, In the image, the object is held in a hand; The importance generating means is a data generating device that generates the importance information using information regarding the size of the hand in the object region. 8. In the data generating device according to any one of claims 1 to 7, The importance generating means is a data generating device that generates the importance information using an importance estimation model that outputs the importance information when the object region is input. 9. In the data generating device according to any one of claims 1 to 8, In the image, the object is held in a hand; The image extraction means is a data generation device that identifies the object region using the position of the hand in the image. 10. In the data generating device according to any one of claims 1 to 9, a surface of the object having a code; The image extraction means is a data generation device that identifies the object region using the position of the code in the image. 11. In the data generating device according to any one of claims 1 to 10, the object is a commodity, The object estimation model is a data generating device used when a product registration device recognizes a product to be registered. 12. A data generation method for generating learning data for generating an object estimation model that estimates the type of an object included in an image, comprising: The computer an image capture process for capturing a plurality of images including the object; an image extraction process for extracting an object region including an object from each of the plurality of images; an importance information generation process for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generation process for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; A data generation method for performing the above. 13. In the data generation method according to claim 12, A data generation method in which, in the image acquisition process, the computer acquires a plurality of frame images included in a single video as at least a portion of the plurality of images. 14. The data generation method according to claim 13, A data generation method in which, in the image extraction process, the computer uses the position of the object area in a first frame image to extract the object area from a second frame image that is generated after the first frame image. 15. In the data generation method according to any one of claims 12 to 14, The data generation method includes performing a model generation process to generate the object estimation model using the training data. 16. In the data generation method according to any one of claims 12 to 15, A data generation method in which, in the importance generation process, the computer generates the importance information using the sharpness of the object in the object region. 17. In the data generation method according to any one of claims 12 to 16, A data generation method in which, in the importance generation process, the computer generates the importance information using information on the size of the object in the object region. 18. In the data generation method according to any one of claims 12 to 17, In the image, the object is held in a hand; A data generation method in which, in the importance generation process, the computer generates the importance information using information about the size of the hand in the object region. 19. In the data generation method according to any one of claims 12 to 18, A data generation method, wherein in the importance generation process, the computer generates the importance information using an importance estimation model that outputs the importance information when the object region is input. 20. In the data generation method according to any one of claims 12 to 19, In the image, the object is held in a hand; A data generation method in which, in the image extraction process, the computer identifies the object region using the position of the hand in the image. 21. In the data generation method according to any one of claims 12 to 20, a surface of the object having a code; A data generation method in which, in the image extraction process, the computer identifies the object region using the position of the code in the image. 22. In the data generation method according to any one of claims 12 to 21, the object is a commodity, The object estimation model is a data generation method used when a product registration device recognizes a product to be registered. 23. A program for causing a computer to function as a data generation device for generating learning data for generating an object estimation model for estimating the type of an object included in an image, comprising: The computer includes: an image capture function for capturing a plurality of images including the object; an image cropping function for cropping an object region including an object from each of the plurality of images; an importance information generating function that generates importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating function for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; A program that allows you to have 24. In the program according to 23 above, The image acquisition function is a program that acquires multiple frame images included in a single video as at least a part of the multiple images. 25. In the program according to 24 above, The image cutting-out function is a program that uses the position of the object area in the first frame image to cut out the object area from a second frame image that is after the first frame image. 26. In the program according to any one of the above 23 to 25, A program for providing the computer with a model generation function for generating the object estimation model using the training data. 27. In the program according to any one of the above 23 to 26, The importance generation function is a program that generates the importance information using the sharpness of the object in the object area. 28. In the program according to any one of the above 23 to 27, The importance generation function is a program that generates the importance information using information regarding the size of the object in the object region. 29. In the program according to any one of the above 23 to 28, In the image, the object is held in a hand; The importance generation function is a program that generates the importance information using information regarding the size of the hand in the object region. 30. In the program according to any one of the above 23 to 29, The importance generation function is a program that generates the importance information using an importance estimation model that outputs the importance information when the object region is input. 31. In the program according to any one of the above 23 to 30, In the image, the object is held in a hand; The image cropping function is a program that identifies the object area using the position of the hand in the image. 32. In the program according to any one of the above 23 to 31, a surface of the object having a code; The image extraction function is a program for identifying the object region using the position of the code in the image. 33. In the program according to any one of the above 23 to 32, the object is a commodity, The object estimation model is a program used by the product registration device when recognizing the product to be registered. [Explanation of symbols]

[0052] 10 Data generation device 20 Object recognition device 110 Image acquisition unit 120 Image extraction section 122 Tracking Department 130 Importance generation part 132 Estimation model memory unit 140 Learning Data Generation Unit 150 Learning data storage unit 160 Model Generation Unit 170 Model Memory Section 180 Transmitter

Claims

1. A data generation device that generates learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: image capture means for capturing a plurality of images including the object; an image extraction means for extracting an object region including an object from each of the plurality of images; an importance generating means for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating means for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Equipped with the image acquisition means acquires a plurality of frame images included in one moving image as at least a portion of the plurality of images; A data generating device, wherein the image cutting means cuts out the object area from a second frame image that is generated after the first frame image, using the position of the object area in a first frame image.

2. A data generation device for generating learning data for generating an object estimation model for estimating a type of an object included in an image, comprising: image capture means for capturing a plurality of images including the object; an image extraction means for extracting an object region including an object from each of the plurality of images; an importance generating means for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating means for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Equipped with A data generating device, wherein the importance generating means generates the importance information using a sharpness of the object in the object region.

3. A data generation device for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: image capture means for capturing a plurality of images including the object; an image extraction means for extracting an object region including an object from each of the plurality of images; an importance generating means for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating means for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Equipped with In the image, the object is held in a hand; The importance generation means generates the importance information using information regarding a size of the hand in the object region.

4. A data generation device for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: image capture means for capturing a plurality of images including the object; an image extraction means for extracting an object region including an object from each of the plurality of images; an importance generating means for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating means for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Equipped with In the image, the object is held in a hand; The image extraction means identifies the object region using the position of the hand in the image.

5. A data generation device for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: image capture means for capturing a plurality of images including the object; an image extraction means for extracting an object region including an object from each of the plurality of images; an importance generating means for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating means for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Equipped with a surface of the object having a code; A data generating device, wherein the image extraction means identifies the object region using the position of the code in the image.

6. The data generating device according to any one of claims 1 to 5, A data generating device comprising: a model generating means for generating the object estimation model using the training data.

7. The data generating device according to any one of claims 1 to 6, The importance generating means is a data generating device that generates the importance information using information regarding the size of the object in the object region.

8. The data generating device according to any one of claims 1 to 7, The importance generating means is a data generating device that generates the importance information using an importance estimation model that outputs the importance information when the object region is input.

9. The data generating device according to any one of claims 1 to 8, the object is a commodity, The object estimation model is a data generating device used when a product registration device recognizes a product to be registered.

10. 1. A data generation method for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: The computer an image capture process for capturing a plurality of images including the object; an image extraction process for extracting an object region including an object from each of the plurality of images; an importance information generation process for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generation process for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Do the following: The computer, In the image acquisition process, a plurality of frame images included in one moving image are acquired as at least a portion of the plurality of images; A data generating method, wherein in the image extraction process, the object area is extracted from a second frame image that is generated after the first frame image, using the position of the object area in a first frame image.

11. A data generation method for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: The computer an image capture process for capturing a plurality of images including the object; an image extraction process for extracting an object region including an object from each of the plurality of images; an importance information generation process for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generation process for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Do the following: The computer, A data generating method, wherein in the importance information generating process, the importance information is generated using a sharpness of the object in the object region.

12. A data generation method for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: The computer an image capture process for capturing a plurality of images including the object; an image extraction process for extracting an object region including an object from each of the plurality of images; an importance information generation process for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generation process for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Do the following: In the image, the object is held in a hand; The computer, A data generation method, wherein in the importance information generation process, the importance information is generated using information on a size of the hand in the object region.

13. A data generation method for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: The computer an image capture process for capturing a plurality of images including the object; an image extraction process for extracting an object region including an object from each of the plurality of images; an importance information generation process for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generation process for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Do the following: In the image, the object is held in a hand; The computer, The data generation method, in the image extraction process, identifying the object region using the position of the hand in the image.

14. A data generation method for generating learning data for generating an object estimation model that estimates a type of an object included in an image, comprising: The computer an image capture process for capturing a plurality of images including the object; an image extraction process for extracting an object region including an object from each of the plurality of images; an importance information generation process for generating importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generation process for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Do the following: a surface of the object having a code; The computer, A data generation method, wherein in the image extraction process, the object region is identified using the position of the code in the image.

15. A program for causing a computer to function as a data generation device for generating learning data for generating an object estimation model that estimates a type of an object included in an image, the program comprising: The computer includes: an image capture function for capturing a plurality of images including the object; an image cropping function for cropping an object region including an object from each of the plurality of images; an importance information generating function that generates importance information indicating an importance of each of the object regions when generating an object estimation model by processing each of the plurality of object regions; a learning data generating function for storing the plurality of object regions and the plurality of pieces of importance information in a storage means as at least a part of the learning data; Let them have The image acquisition function acquires a plurality of frame images included in one moving image as at least a portion of the plurality of images, The image cropping function is a program that crops out the object area from a second frame image that is subsequent to the first frame image, using the position of the object area in a first frame image.

Images