Information processing method, model generation method, program, and information processing apparatus

Abstract

To provide an information processing method and the like capable of determining authenticity of a wristwatch with high accuracy.SOLUTION: An information processing method according to an embodiment of the present disclosure includes acquiring a dial plate image of a wristwatch, acquiring a model of the wristwatch by inputting the acquired dial plate image to a first learning model that outputs the model of the wristwatch in a case where the dial plate image of the wristwatch is input, specifying a recognition region corresponding to the acquired model, acquiring a region image of the wristwatch corresponding to the recognition region, and determining authenticity of the wristwatch by inputting the acquired region image to a second learning model used for determining authenticity of the wristwatch.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present technology relates to an information processing method, a model generation method, a program, and an information processing device. [Background technology]

[0002] Conventionally, information processing systems for acquiring item identification information for identifying an item have been proposed. For example, the information processing system described in Patent Document 1 includes an item group information acquisition unit that acquires item group information related to an item group to which an item belongs from a first image obtained by capturing a first side of the item, an item feature amount generation unit that generates feature amounts of the item from a second image obtained by capturing a second side of the item, and an item identification information acquisition unit that acquires item identification information for identifying the item based on the item group information and the feature amounts. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7147087 Summary of the Invention [Problem to be solved by the invention]

[0004] Counterfeit wristwatches are in circulation. In recent years, the accuracy with which counterfeits resemble genuine watches has increased, so the technology described in Patent Document 1 cannot accurately determine the authenticity of wristwatches.

[0005] The present disclosure has been made in consideration of the above circumstances, and aims to provide an information processing method and the like that can determine the authenticity of a wristwatch with high accuracy. [Means for solving the problem]

[0006] An information processing method according to one embodiment of the present disclosure acquires a watch face image, inputs the acquired watch face image into a first learning model that outputs the model of the watch when the watch face image is input to acquire the model of the watch, identifies a recognition area corresponding to the acquired model, acquires an area image of the watch that corresponds to the recognition area, and inputs the acquired area image into a second learning model used to determine the authenticity of the watch, thereby determining the authenticity of the watch. [Effects of the Invention]

[0007] With the information processing method according to one embodiment of the present disclosure, it is possible to determine the authenticity of a wristwatch with high accuracy. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is an explanatory diagram illustrating an example of an information processing system. [Figure 2] FIG. 1 is a block diagram illustrating an example of an information processing device. [Figure 3] FIG. 2 is a block diagram illustrating an example of a terminal device. [Figure 4] FIG. 10 is an explanatory diagram illustrating an example of a model table. [Figure 5] FIG. 10 is an explanatory diagram illustrating an example of a type determination model. [Figure 6] FIG. 10 is an explanatory diagram illustrating an example of an authenticity determination model. [Figure 7] FIG. 10 is an explanatory diagram showing an example of a type determination screen. [Figure 8] FIG. 10 is an explanatory diagram showing an example of an authenticity determination model learning screen. [Figure 9] FIG. 10 is an explanatory diagram showing an example of an authenticity determination screen. [Figure 10] FIG. 10 is an explanatory diagram showing an example of a determination result display screen. [Figure 11] 10 is a flowchart illustrating an example of a learning process of an authenticity determination model. [Figure 12] 10 is a flowchart illustrating an example of an authenticity determination process. [Figure 13]FIG. 10 is an explanatory diagram showing an example of an authenticity determination screen according to a modified example. [Figure 14] FIG. 10 is an explanatory diagram showing an example of a determination result display screen according to a modified example. DETAILED DESCRIPTION OF THE INVENTION

[0009] (Embodiment) FIG. 1 is an explanatory diagram showing an example of an information processing system S. The information processing system S includes an information processing device 1 and a terminal device 2. The information processing device 1 is, for example, a server, and stores various tables and learning models. The terminal device 2 is, for example, a smartphone. The terminal device 2 may also be a wearable computer, smart glasses, a tablet terminal, a personal computer, or the like. The terminal device 2 photographs a wristwatch and transmits the photographed image (dial image and area image) to the information processing device 1. Based on the image acquired from the terminal device 2, the information processing device 1 recognizes (determines) whether the photographed wristwatch is genuine or a counterfeit, and displays the determination result on the terminal device 2.

[0010] FIG. 2 is a block diagram showing an example of an information processing device 1. The information processing device 1 includes a processing unit 11, a storage unit 12, and a communication unit 13. The processing unit 11 includes a central processing unit (CPU), a microprocessing unit (MPU), a graphics processing unit (GPU), a neural processing unit (NPU), or a tensor processing unit (TPU), and performs various information processing, control processing, and the like related to the information processing device 1 by reading and executing a program P (a program product) stored in the storage unit 12. The program P can be deployed to be executed on a single computer or at a single site, or distributed across multiple sites and on multiple computers interconnected by a communication network. The information processing device 1 may also be a device corresponding to a node on a blockchain. While FIG. 1 illustrates the processing unit 11 as a single processor, it may also be a multiprocessor.

[0011] The storage unit 12 of the information processing device 1 is, for example, a volatile memory and a non-volatile memory. The storage unit 12 stores a program P, a type table 121, a type determination model (first learning model) M1, and an authenticity determination model (second learning model) M2. The program P may be provided to the information processing device 1 using a computer-readable storage medium 12a. The storage medium 12a may be, for example, a portable memory. Examples of the portable memory include a CD-ROM, a USB memory, an SD card, a microSD card, and a Compact Flash Memory (registered trademark). When the storage medium 12a is a portable memory, the processing element of the processing unit 11 may read the program P from the storage medium 12a using a reading device (not shown). The read program P is written to the storage unit 12. Furthermore, the program P may be provided to the information processing device 1 by the communication unit 13 communicating with an external device. Details of the type table 121, the type determination model M1, and the authenticity determination model M2 will be described later. The model table 121 may be stored in a database server different from the information processing device 1.

[0012] The communication unit 13 of the information processing device 1 is a communication module or communication interface for communicating with other devices such as the terminal device 2 by wire or wirelessly, and is, for example, a wide-area wireless communication module such as LTE (registered trademark), 4G, or 5G. The processing unit 11 communicates with the terminal device 2 via the communication unit 13 through an external network N such as the Internet.

[0013] FIG. 3 is a block diagram showing an example configuration of the terminal device 2. The terminal device 2 includes a device processing unit 21, a storage unit 22, a communication unit 23, a display unit 24, an input unit 25, and a photographing unit 26. The device processing unit 21 is configured with a central processing unit (CPU), a micro processing unit (MPU), a graphics processing unit (GPU), a neural processing unit (NPU), a tensor processing unit (TPU), or the like, and performs various control processes, arithmetic processes, and the like. Note that the functions of the terminal device 2 may be realized by multiple devices, or may be a device corresponding to a node on a blockchain. Furthermore, the terminal device 2 may perform some or all of the processes executed by the information processing device 1.

[0014] The storage unit 22 of the terminal device 2 stores an application program (app program) Pa that transmits photographed images of the wristwatch to the information processing device 1 and displays the results of authentication of the wristwatch received from the information processing device 1. The app program Pa is provided to the terminal device 2 using, for example, a recording medium 22a. The recording medium 22a is, for example, a portable memory. Examples of portable memory include a CD-ROM, a USB (Universal Serial Bus) memory, an SD card, a microSD card, and a CompactFlash Memory (registered trademark). When the recording medium 22a is a portable memory, the processing element of the device processing unit 21 may read the app program Pa from the recording medium 22a using a reading device (not shown). The device processing unit 21 of the terminal device 2 may obtain the app program Pa via the Internet and store it in the storage unit 22. The storage unit 22 of the terminal device 2 may also store a type determination model M1 or an authenticity determination model M2, and the device processing unit 21 of the terminal device 2 may perform processing using the type determination model M1 or the authenticity determination model M2.

[0015] The communication unit 23 of the terminal device 2 is a communication module for communicating with the information processing device 1 via a wired or wireless network N. The terminal processing unit 21 communicates with the information processing device 1 via the communication unit 23 and the external network N.

[0016] The display unit 24 of the terminal device 2 is composed of a display and displays a screen for sending a captured image to the information processing device 1, or a screen for displaying the authenticity determination results obtained from the information processing device 1, etc.

[0017] The input unit 25 of the terminal device 2 accepts input of various information or input of selection. When the terminal device 2 is a smartphone, the display unit 24 and the input unit 25 are integrally configured as a touch panel.

[0018] The image capturing unit 26 of the terminal device 2 captures images of the watch face, area images, and learning videos. The image capturing unit 26 according to this embodiment is realized by a camera built into a smartphone. Note that the image capturing unit 26 may also be realized by a camera connected to the terminal device 2 by wire or wirelessly.

[0019] 4 is an explanatory diagram showing an example of the model table 121. The model table 121 stores information related to each model, as well as the recognition area and authenticity determination model corresponding to the model. The management items (fields) of the model table 121 include a model field, a manufacturer field, a model name field, a model name field, a recognition area field, and an authenticity determination model name field.

[0020] The model field of model table 121 stores a number (model ID) assigned to each model of wristwatch. The manufacturer field stores the manufacturer name of the wristwatch corresponding to the model, such as ROLEX (registered trademark) or OMEGA (registered trademark). The model name field stores the model name of the wristwatch corresponding to the model, such as DAYTONA (registered trademark), SUBMARINER (registered trademark), or SPEEDMASTAR (registered trademark). The model name field stores the model name of the wristwatch corresponding to the model.

[0021] The recognition area field of the model table 121 stores information about the part of the watch that the user is asked to photograph (recognition area) corresponding to the model. The recognition area may be, for example, a logo, belt, buckle, crown, dial, back of the dial, or side. After recognizing the model of the photographed watch, the processing unit 11 of the information processing device 1 transmits information requesting the user to photograph the part of the watch stored in the recognition area field to the terminal device 2. The authenticity determination model name field stores a number (code) that identifies the authenticity determination model M2 corresponding to the identified model of the watch. In other words, the memory unit 12 stores an authenticity determination model M2 for each model. The processing unit 11 of the information processing device 1 inputs the acquired image of the recognition area (area image) into the authenticity determination model M2 stored in the authenticity determination model name field to determine the authenticity of the watch.

[0022] 5 is an explanatory diagram showing an example of the type determination model M1. The type determination model M1 is a model with an object detection function, such as a CNN (Convolutional Neural Network), RCNN (Regions with Convolutional Neural Network), Fast RCNN, Faster RCNN, SSD (Single Shot Multibook Detector), YOLO (You Only Look Once), or Vision Transformer. The type determination model M1 is trained to output the type of watch when a watch face image is input, using training data that previously associates watch face images with watch models.

[0023] If the type determination model M1 is configured with a neural network including a CNN that extracts image features, such as RCNN, the type determination model M1 has multiple neurons that accept input of pixel values ​​of the watch face image and passes the input pixel values ​​to the intermediate layer. The intermediate layer has multiple neurons that extract image features of the watch face image and passes the extracted image features to the output layer. The output layer outputs the type ID of the watch photographed in the watch face image based on the image features. Note that the processing unit 11 may also output the type ID using a technique such as pattern matching. If the watch photographed in the watch face image is genuine, the type determination model M1 outputs the type ID of the watch model. If the watch photographed in the watch face image is a counterfeit, the type determination model M1 outputs the type ID of the watch model that the counterfeit copy was made of.

[0024] FIG. 6 is an explanatory diagram showing an example of the authenticity determination model M2. The authenticity determination model M2 is, for example, an autoencoder. The authenticity determination model M2 has an encoder, a decoder, a comparator, and a determiner. When an area image of a wristwatch's recognition area is input, the encoder of the authenticity determination model M2 extracts latent variables from the area image by reducing the dimension of the area image data. The decoder of the authenticity determination model M2 generates a generated area image based on the extracted latent variables. The comparator of the authenticity determination model M2 calculates the probability that the photographed wristwatch is genuine or a counterfeit based on the difference between the area image and the generated area image. The smaller the difference, the higher the probability that the wristwatch is genuine and the lower the probability that it is a counterfeit; and the larger the difference, the lower the probability that the wristwatch is genuine and the higher the probability that it is a counterfeit. At this time, the processing unit 11 outputs the probability that the wristwatch is genuine or a counterfeit by calculating the difference in contrast or color between the region image and the generated region image and the mean square value of the difference. The determiner of the authenticity determination model M2 determines that the wristwatch is genuine if the probability of authenticity is a predetermined value (e.g., 50%) or higher, and determines that the wristwatch is a counterfeit if the probability of authenticity is less than the predetermined value, and outputs the determination result. The authenticity determination model M2 may output the wristwatch authenticity determination result by outputting the area in which a genuine or counterfeit wristwatch is displayed using semantic segmentation, instance segmentation, or panoptic segmentation. In this embodiment, the authenticity determination model M2 is configured with 50 convolutional layers, but it may also be configured with 3 to 152 convolutional layers, or with 152 or more convolutional layers.

[0025] FIG. 7 is an explanatory diagram showing an example of a type determination screen. A circular dial guide and a capture button are displayed on the type determination screen. In FIG. 7, the dial guide is indicated by a dashed circle. The user can use the capture unit 26 to align the dial guide with the watch dial and press the capture button to capture a dial image. The device processing unit 21 of the terminal device 2 transmits the captured dial image to the information processing device 1. Note that the device processing unit 21 of the terminal device 2 may transmit to the information processing device 1 a dial image stored in the memory unit 22 or an image in which the dial area is cropped from the image of the watch.

[0026] The processing unit 11 of the information processing device 1 that has acquired the watch face image inputs the watch face image into the type determination model M1 to acquire the type ID (type) of the wristwatch whose watch face was photographed. If the type determination model M1 is unable to output the type of the wristwatch (unable to determine the type), the processing unit 11 of the information processing device 1 may cause the terminal device 2 to accept input of the manufacturer, model name, and type name of the wristwatch, and acquire the input manufacturer, model name, and type name. At this time, the processing unit 11 of the information processing device 1 may generate a new type ID corresponding to the input manufacturer, model name, and type name, associate the generated type ID with the manufacturer, model name, and type name, store them in the type table 121, and re-train the type determination model M1 using training data that associates watch faces with type IDs.

[0027] 8 is an explanatory diagram showing an example of an authenticity determination model learning screen. The authenticity determination model learning screen displays a type display field, a recognition area guide, and a video recording button. The type display field on the authenticity determination model learning screen displays the manufacturer, model name, and type name of the wristwatch read from the type table 121 that corresponds to the type obtained by the processing unit 11 of the information processing device 1 using the type determination model M1.

[0028] Above the recognition area guide on the authenticity determination model learning screen, the recognition area of ​​the wristwatch to be photographed, which corresponds to the model and is read from the model table 121, is displayed. In FIG. 8, the recognition area guide is indicated by a dashed rectangle. The user can use the photographing unit 26 to align the recognition area of ​​the wristwatch with the recognition area guide in accordance with the displayed recognition area, and then press the video recording button to shoot a learning video. The wristwatch photographed on the authenticity determination model learning screen is an authentic wristwatch. The learning video is filmed for a predetermined period of time (e.g., 20 seconds). The device processing unit 21 of the terminal device 2 transmits the filmed learning video to the information processing device 1. The device processing unit 21 may also transmit a video stored in the memory unit 22 to the information processing device 1 as the learning video.

[0029] When the processing unit 11 of the information processing device 1 acquires region images (learning video) of a predetermined time, it performs data expansion of the multiple region images included in the learning video into a larger number of region images. Data expansion is performed, for example, by pixel shifting, angle rotation, enlargement, or reduction. Region image data expansion may also be performed manually. For example, if the learning video is shot at 60 fps for 20 seconds, the processing unit 11 performs data expansion from 1,200 region images to approximately 30,000 to 50,000 region images. Note that the number of region images expanded by the processing unit 11 is not limited to this. When a region image is input, the processing unit 11 trains the authenticity determination model M2 so that there is no difference between the input region image and a generated region image generated based on the input region image. Note that if the learning video is a photograph of a counterfeit wristwatch, the processing unit 11 may train the authenticity determination model M2 so that there is a larger difference between the region image and the generated region image. Furthermore, the terminal processing unit 21 of the terminal device 2 may capture a learning video relating to a plurality of recognition regions and transmit it to the information processing device 1. At this time, the processing unit 11 of the information processing device 1 may train the authenticity determination model M2 based on region images of the learning video relating to the plurality of recognition regions.

[0030] 9 is an explanatory diagram showing an example of the authenticity determination screen. The authenticity determination screen includes a type display field, a recognition area guide, and an area photographing button. The type display field of the authenticity determination screen displays the manufacturer, model name, and type name of the wristwatch read from the type table 121, which correspond to the type acquired by the processing unit 11 of the information processing device 1 using the type determination model M1.

[0031] Above the recognition area guide on the authenticity determination screen, the recognition area of ​​the wristwatch to be photographed, which corresponds to the model and is read from model table 121, is displayed. In FIG. 9, the recognition area guide is indicated by a dashed rectangle. The user can use the photographing unit 26 to align the wristwatch recognition area with the recognition area guide in accordance with the displayed recognition area, and then press the photographing button to photograph an area image. Note that the device processing unit 21 may transmit an image stored in the memory unit 22 to the information processing device 1 as an area image.

[0032] When the processing unit 11 of the information processing device 1 acquires the region image, it inputs the acquired region image into the authenticity determination model M2 corresponding to the model of the wristwatch and outputs the probability that the wristwatch is genuine. For example, if the probability that the wristwatch is genuine is equal to or greater than a predetermined value, the processing unit 11 determines that the wristwatch is genuine. If the probability that the wristwatch is genuine is less than the predetermined value, the processing unit 11 determines that the wristwatch is a counterfeit. The processing unit 11 outputs the probability that the wristwatch is genuine and the determination result to the terminal device 2.

[0033] 10 is an explanatory diagram showing an example of the judgment result display screen. The judgment result display screen displays the probability that the wristwatch is genuine, as received from the information processing device 1 by the device processing unit 21 of the terminal device 2, and the judgment result. The judgment result display screen also displays the manufacturer, model name, and type name of the wristwatch. The judgment result display screen may also display the probability that the wristwatch is a counterfeit.

[0034] 11 is a flowchart illustrating an example of the learning process of the authenticity determination model M2. The device processing unit 21 of the terminal device 2 displays a type determination screen on the display unit 24 (S1). The device processing unit 21 captures a dial image (S2) and transmits the captured dial image to the information processing device 1 (S3).

[0035] The processing unit 11 of the information processing device 1 inputs the watch face image acquired from the terminal device 2 into the type determination model M1 (S4) and outputs the type ID of the wristwatch (S5). The processing unit 11 uses the type ID as a key to read out the manufacturer, model name, type name, and recognition area corresponding to the type of the wristwatch from the type table 121 (S6). The processing unit 11 transmits the read out manufacturer, model name, type name, and recognition area to the terminal device 2 (S7).

[0036] The device processing unit 21 of the terminal device 2 displays an authenticity determination model learning screen on the display unit 24, which displays the manufacturer, model name, type name, and recognition area received from the information processing device 1 (S8). The device processing unit 21 then uses the shooting unit 26 to capture a learning video showing the recognition area of ​​the wristwatch (S9), transmits the captured learning video to the information processing device 1 (S10), and ends the process.

[0037] The processing unit 11 of the information processing device 1 performs data augmentation (augmentation) on multiple area images included in the learning video acquired from the terminal device 2 to further increase the number of area images (S11). The processing unit 11 inputs the data-augmented multiple area images to the authenticity determination model M2 (S12), trains the authenticity determination model M2 so that the differences between the area images and the generated area images are eliminated (S13), and ends the process. Wristwatches, which need to be judged as genuine or counterfeit, are expensive and difficult to obtain. Therefore, the number of area images taken manually is small, and it has been difficult to secure a sufficient number of area images for sufficient training of the authenticity determination model M2. By capturing a video of the watch and further performing data augmentation based on each frame included in the video, a sufficient number of area images can be input to train the authenticity determination model M2.

[0038] Fig. 12 is a flowchart illustrating an example of the authenticity determination process. The processes in S21 to S27 are the same as those in S1 to S7 shown in Fig. 11. The device processing unit 21 of the terminal device 2 displays an authenticity determination screen on the display unit 24, which displays the manufacturer, model name, type name, and recognition area received from the information processing device 1 (S28). The device processing unit 21 then uses the photographing unit 26 to photograph an area image showing the recognition area of ​​the wristwatch (S29), and transmits the photographed area image to the information processing device 1 (S30).

[0039] The processing unit 11 of the information processing device 1 inputs the area image acquired from the terminal device 2 to the encoder of the authenticity determination model M2 (S31) and extracts latent variables of the area image (S32). The processing unit 11 outputs a generated area image using the decoder of the authenticity determination model M2 based on the latent variables (S33). The processing unit 11 calculates the difference between the area image and the generated area image (S34), and outputs the probability that the wristwatch is authentic based on the calculated difference (S35). The processing unit 11 determines whether the wristwatch is authentic or a counterfeit (authenticity) based on the probability that the wristwatch is authentic (S36). The processing unit 11 transmits the probability that the wristwatch is authentic and the determination result in S36 to the terminal device 2 (S37), and ends the processing.

[0040] The terminal processing unit 21 of the terminal device 2 displays on the display unit 24 the probability that the wristwatch received from the information processing device 1 is genuine and a determination result display screen showing the determination result (S38), and then ends the process.

[0041] (Variation) In the embodiment, one recognition area corresponds to one model, but two may also be provided. In a modified example, a first recognition area and a second recognition area are identified for the model acquired by the processing unit 11 of the information processing device 1. Note that three or more recognition areas may also be provided for one model.

[0042] 13 is an explanatory diagram showing an example of an authenticity determination screen according to a modified example. The authenticity determination screen according to the modified example displays a type display field, a first recognition area guide, a first area photographing button, a second recognition area guide, and a second area photographing button. The type display field on the authenticity determination screen displays the manufacturer, model name, and type name of the wristwatch read from the type table 121 that correspond to the type acquired by the processing unit 11 of the information processing device 1 using the type determination model M1.

[0043] The model table 121 according to the modified example stores first and second recognition areas corresponding to the model. The first recognition area of ​​the wristwatch to be photographed, which corresponds to the model and is read from the model table 121, is displayed above the first recognition area guide on the authenticity determination screen. The user can use the photographing unit 26 to align the first recognition area of ​​the wristwatch with the first recognition area guide in accordance with the displayed recognition area, and then press the first photographing button to photograph the first area image. The device processing unit 21 may also transmit an image stored in the memory unit 22 to the information processing device 1 as the first area image.

[0044] Above the second recognition area guide on the authenticity determination screen, a second recognition area of ​​the wristwatch to be photographed, which corresponds to the model and is read from model table 121, is displayed. The user can use the photographing unit 26 to align the second recognition area of ​​the wristwatch with the second recognition area guide in accordance with the displayed recognition area, and then press the second photographing button to photograph a second area image. Note that the device processing unit 21 may transmit an image stored in the memory unit 22 to the information processing device 1 as the second area image. The second recognition area guide and second photographing button may also be displayed on a second authenticity determination screen that is displayed after the first area image is photographed on the authenticity determination screen.

[0045] In a modified example, the storage unit 12 of the information processing device 1 stores an authenticity determination model M2 corresponding to each wristwatch model and recognition area (first recognition area and second recognition area). For example, as shown in FIG. 13, if the first recognition area is the crown, the authenticity determination model M2 for the first recognition area of ​​the wristwatch is trained based on a training video of the crown of the wristwatch. If the second recognition area is the buckle, the authenticity determination model M2 for the second recognition area of ​​the wristwatch is trained based on a training video of the buckle. When the processing unit 11 of the information processing device 1 acquires the first and second region images, it inputs the acquired first and second region images into the authenticity determination model M2 corresponding to the wristwatch model and recognition area, and outputs the probability that the wristwatch is authentic for each of the first and second region images. Note that the processing unit 11 may also input the acquired first and second region images into the authenticity determination model M2 corresponding to the wristwatch model. For example, if the probabilities that the wristwatch is genuine for both the first area image and the second area image are equal to or greater than a predetermined value, the processing unit 11 determines that the wristwatch is genuine. If the probabilities that the wristwatch is genuine for at least one of the first area image or the second area image are less than a predetermined value, the processing unit 11 determines that the wristwatch is a counterfeit. The processing unit 11 outputs the probabilities that the wristwatch is genuine for each of the first area image and the second area image, and the determination result, to the terminal device 2.

[0046] 14 is an explanatory diagram showing an example of a determination result display screen according to a modified example. The determination result display screen displays the probability that the wristwatch is genuine for each of the first area image and the second area image acquired by the device processing unit 21 of the terminal device 2 from the information processing device 1, and the determination result. Note that the determination result display screen may display only the lowest probability of authenticity for the first area image or the second area image, or may display an average value of the authenticity probabilities for the first area image and the second area image.

[0047] According to the above configuration and processing, the processing unit 11 of the information processing device 1 determines the model of a wristwatch based on a dial image and determines the authenticity of the wristwatch based on an area image of a recognition area corresponding to the model, thereby enabling highly accurate authentication of the wristwatch. Furthermore, the authenticity determination model M2 is trained using a large number of area images that have been data-expanded based on training videos of genuine wristwatches, thereby easily improving the accuracy of authenticity determination by the authenticity determination model M2. In this embodiment, the authenticity determination model M2 is an autoencoder, but the authenticity determination model M2 may also be CNN, RCNN, Fast RCNN, Faster RCNN, SSD, YOLO, Vision Transformer, or the like. In this case, the authenticity determination model M2 may also be trained using training data that associates area images of a wristwatch with information indicating whether the wristwatch is genuine or a counterfeit. Furthermore, the objects for which authenticity determination is performed are not limited to wristwatches. The processing unit 11 of the information processing device 1 may determine the type and authenticity of products such as bags, wallets, jewelry, accessories, shoes, clothes, or carpets (so-called branded products).

[0048] The embodiments disclosed herein are illustrative in all respects and should not be considered limiting. The technical features described in each embodiment may be combined with one another, and the scope of the present invention is intended to include all modifications within the scope of the claims and equivalents thereto. Furthermore, independent and dependent claims described in the claims may be combined with one another in any and all combinations, regardless of the reference format. Furthermore, while the claims use a format in which a claim references two or more other claims (multiple claim format), this is not limiting. Multiple claims (multiple multiple claims) that reference at least one other multiple claim may also be used.

[0049] (Addendum) Get a picture of the product inputting the acquired image of the product into a first learning model that outputs the model of the product when the image of the product is input, and acquiring the model of the product; Identifying a recognition area corresponding to the acquired model, acquiring an area image of the product corresponding to the recognition area; The acquired area image is input to a second learning model used to recognize the product, thereby recognizing the product. Information processing methods. [Explanation of symbols]

[0050] 1: Information processing equipment 11: Processing section 12: Storage section 121: Model table 12a: Recording medium 13: Communications Department 2: Terminal device 21: Terminal processing section 22: Storage section 22a: Recording medium 23: Communications Department 24: Display section 25: Input section 26: Photography Department M1: Type determination model M2: Authenticity determination model N: Network P: Program Pa: Application program S: Information Processing System

Claims

1. Get the watch face image, inputting the acquired watch face image into a first learning model that outputs the model of a watch when the watch face image is input, and acquiring the model of the watch; Identifying a recognition area corresponding to the acquired model, Acquire an area image of the wristwatch corresponding to the recognition area; The acquired region image is input into a second learning model used to determine the authenticity of the wristwatch, thereby determining the authenticity of the wristwatch. Information processing methods.

2. Output the manufacturer, model name, and type name corresponding to the acquired model. The information processing method according to claim 1 .

3. A table is referenced that stores the manufacturer and model name or type name associated with each model of the wristwatch, and the manufacturer and model name or type name corresponding to the acquired model is output. The information processing method according to claim 2 .

4. outputting the recognition region; Acquire the area image in which the recognition area is captured 3. The information processing method according to claim 1.

5. outputting the first recognition region; acquiring a first area image in which the first recognition area is captured; outputting the second recognition region; A second area image in which the second recognition area is captured is acquired.

3. The information processing method according to claim 1.

6. storing the second learning model for each type; By inputting the region image into the second learning model corresponding to the acquired model, the probability that the wristwatch is a genuine product or a counterfeit product is output.

3. The information processing method according to claim 1.

7. Outputs the manufacturer, model name, and type name corresponding to the acquired model, as well as the probability that the wristwatch is genuine or a counterfeit. The information processing method according to claim 6.

8. The second learning model is an autoencoder, and determines the authenticity of the wristwatch based on an input area image and a generated area image generated by a function of the autoencoder.

3. The information processing method according to claim 1.

9. Get the watch face image, inputting the acquired watch face image into a first learning model that outputs the model of a watch when the watch face image is input, and acquiring the model of the watch; Identifying a recognition area corresponding to the acquired model, Acquire an image of the wristwatch area corresponding to the recognition area for a predetermined period of time; A second learning model is generated based on the acquired plurality of area images to be used for determining the authenticity of the wristwatch. How the model is generated.

10. Get an image of the watch, inputting the acquired image of the wristwatch into a first learning model that outputs the model of the wristwatch when an image of the wristwatch is input, and acquiring the model of the wristwatch; Identifying a recognition area corresponding to the acquired model, Acquire an area image of the wristwatch corresponding to the recognition area; The acquired region image is input into a second learning model used to determine the authenticity of the wristwatch, thereby determining the authenticity of the wristwatch. A program that causes a computer to perform a process.

11. Get an image of the watch, inputting the acquired image of the wristwatch into a first learning model that outputs the model of the wristwatch when an image of the wristwatch is input, and acquiring the model of the wristwatch; Identifying a recognition area corresponding to the acquired model, Acquire an area image of the wristwatch corresponding to the recognition area; The acquired region image is input into a second learning model used to determine the authenticity of the wristwatch, thereby determining the authenticity of the wristwatch. Processing section An information processing device comprising:

Images