Inference program, machine learning program, inference method, machine learning method, and information processing device

Abstract

According to the embodiments, an inference program makes a computer execute acquisition processing and inference processing. The acquisition processing acquires object analysis information that indicates physicochemical analysis results that correspond to a test object by inputting a test image obtained by non-destructive testing of the test object into a first machine learning model that has been trained by machine learning that uses a data set of first training data that associates a test image obtained by non-destructive testing of a first food product and first analysis information that indicates physicochemical analysis results for the first food product. The inference processing infers sensory evaluation results that correspond to the test object by inputting the acquired object analysis information into a second machine learning model that has been trained by machine learning that uses a data set of second training data that associates second analysis information that indicates physicochemical analysis results for a second food product and sensory evaluation results for the second food product.

Description

Inference processing program, machine learning program, inference processing method, machine learning method, and information processing device

[0001] An embodiment of the present invention relates to an inference processing program, a machine learning program, an inference processing method, a machine learning method, and an information processing device.

[0002] In the past, there has been a trend to promote branding of foods such as fruits by quantitatively evaluating their quality and classifying them into grades. In such food quality evaluations, it is desirable to perform non-destructive and simple testing so as not to damage the commercial value of the subject.

[0003] A known conventional technique for non-destructively evaluating the quality of food and other test objects is to measure the sugar and acidity levels inside the food using near-infrared spectroscopy.

[0004] "Near-infrared Non-destructive Vegetable and Fruit Quality Checker F-750," [online], [searched July 27, 2023], Internet <URL: https: / / www.toyokokagaku.co.jp / product / development / felix / 003.html>

[0005] Sensory evaluation using five tastes (sweetness, umami, bitterness, sourness, and saltiness) is a standard method for quantifying the palatability of food. However, the above-mentioned conventional techniques are limited to measuring sugar content and acidity, and do not involve sensory evaluation using the five tastes (sweetness, umami, bitterness, sourness, and saltiness). Furthermore, sensory evaluation is generally performed by an inspector tasting the food to be inspected, and in practice, it is difficult to perform non-destructive and simple evaluation due to the risk of some food being lost due to sampling inspection and the difficulty of securing skilled inspectors.

[0006] In one aspect, an object of the present invention is to provide an inference processing program, a machine learning program, an inference processing method, a machine learning method, and an information processing device that can non-destructively and easily obtain a sensory evaluation of a test object.

[0007] In one proposal, the inference processing program causes a computer to execute an acquiring process and an inference process. The acquiring process acquires target analytical information indicating the physicochemical analysis results corresponding to the test target by inputting an inspection image obtained by non-destructive testing of a first food to a first machine learning model trained by machine learning using a first training data dataset that associates an inspection image obtained by non-destructive testing of a first food with first analytical information indicating the physicochemical analysis results of the first food. The inference process infers the sensory evaluation results corresponding to the test target by inputting the acquired target analytical information to a second machine learning model trained by machine learning using a second training data dataset that associates second analytical information indicating the physicochemical analysis results of a second food with the sensory evaluation results of the second food.

[0008] According to one embodiment, a sensory evaluation of a test object can be obtained non-destructively and simply.

[0009] FIG. 1 is an explanatory diagram illustrating an overview of processing of an information processing device according to an embodiment. FIG. 2 is a block diagram illustrating an example of the functional configuration of an information processing device according to an embodiment. FIG. 3 is a flowchart illustrating an example of operation of an information processing device according to an embodiment. FIG. 4 is an explanatory diagram illustrating learning of an autoencoder. FIG. 5 is an explanatory diagram illustrating learning of a first machine learning model. FIG. 6 is an explanatory diagram illustrating learning of a second machine learning model. FIG. 7 is an explanatory diagram illustrating inference of evaluation results. FIG. 8 is an explanatory diagram illustrating an example of a computer configuration.

[0010] Hereinafter, an inference processing program, a machine learning program, an inference processing method, a machine learning method, and an information processing device according to embodiments will be described with reference to the drawings. Components having the same functions in the embodiments will be assigned the same reference numerals, and duplicate descriptions will be omitted. Note that the inference processing program, the machine learning program, the inference processing method, the machine learning method, and the information processing device described in the following embodiments are merely examples and do not limit the embodiments. Furthermore, the following embodiments may be combined as appropriate within a consistent range.

[0011] 1 is an explanatory diagram illustrating an overview of processing performed by an information processing device according to an embodiment. As shown in Fig. 1, the information processing device according to the embodiment infers an evaluation result 103 of a sensory evaluation regarding five tastes (sweetness, umami, bitterness, sourness, and saltiness) based on inspection image data 100 of an inspection image (an ultrasound image in the illustrated example) obtained by non-destructive testing of an inspection object such as food.

[0012] In the embodiment described below, an example is given in which an ultrasound image obtained by ultrasound inspection of an object to be inspected using an inspection probe or the like is used as the inspection image, but the type of inspection image is not limited to an ultrasound image. For example, the inspection image may be a near-infrared image of the object to be inspected using near-infrared spectroscopy, a radiological image obtained by irradiating the object to be inspected with radiation, or a combination of these various inspection images. Furthermore, the ultrasound image may also be various mode images such as an A (amplitude) mode image and a B (brightness) mode image, or a combination of these various mode images.

[0013] The information processing device according to the embodiment inputs inspection image data 100 obtained by non-destructive inspection of an inspection object into a first machine learning model 110, thereby obtaining physicochemical information 102 indicating the results of physicochemical analysis corresponding to the inspection object.

[0014] The first machine learning model 110 is a model trained by supervised machine learning using a first training dataset for each food example, so that when test image data 100 of the test object is input, it outputs physicochemical information 102 corresponding to the test object.

[0015] Specifically, the first training data set includes a data set for each case that combines inspection image data of an inspection image obtained by non-destructive testing of food with physicochemical analysis data 101 that indicates the results of physicochemical analysis of the food.

[0016] The physicochemical analysis data 101 is multidimensional data that includes, as physicochemical analysis results of food, chemical analysis results such as salt, water, fat, and protein content, as well as physical analysis results such as hardness and mass.

[0017] If machine learning of the first machine learning model 110 is performed using such multidimensional physicochemical analysis data 101, there is a risk of overlearning, which results in excessive adaptation to the physicochemical analysis data 101. Therefore, the information processing device according to the embodiment performs dimensional compression of the physicochemical analysis data 101 included in the first training data set using an autoencoder 111 to generate Embedding (physicochemical information 102). The physicochemical information 102 generated in this manner is data in which feature quantities of the physicochemical analysis results included in the physicochemical analysis data 101 are expressed using lower-dimensional vector representations or the like.

[0018] When inspection image data of an inspection image obtained by non-destructive inspection of food is input to the first machine learning model 110, the information processing device according to the embodiment performs machine learning on the first machine learning model 110 so that the output from the first machine learning model 110 becomes physicochemical information 102 generated from physicochemical analysis data 101. Specifically, the information processing device according to the embodiment adjusts parameters of the first machine learning model 110 using a known machine learning algorithm such as backpropagation.

[0019] After the physicochemical information 102 is obtained, the information processing device according to the embodiment inputs the physicochemical information 102 into a second machine learning model 112, thereby inferring an evaluation result 103 of the sensory evaluation corresponding to the test object.

[0020] The second machine learning model 112 is a model trained by supervised machine learning using a second training dataset for each food example, to input physicochemical information 102 obtained from test image data 100 of the test object and output a sensory evaluation result 103 corresponding to the test object.

[0021] Specifically, the second training data set includes a data set for each case that combines physicochemical analysis data 101 indicating the results of physicochemical analysis of food with the evaluation results of sensory evaluation of the food.

[0022] The information processing device according to the embodiment performs dimensional compression of physicochemical analysis data 101 included in a second training dataset using an autoencoder 111 to generate Embedding (physicochemical information 102). Next, the information processing device according to the embodiment performs machine learning on the second machine learning model 112 so that when the generated physicochemical information 102 is input to the second machine learning model 112, the output from the second machine learning model 112 becomes the evaluation result of the sensory evaluation of the food. Specifically, like the first machine learning model 110, the information processing device according to the embodiment adjusts the parameters of the second machine learning model 112 using a known machine learning algorithm such as backpropagation.

[0023] In the information processing device according to the embodiment, the sensory evaluation result 103 can be inferred based on the inspection image data 100 obtained by non-destructive testing of the inspection object, thereby making it possible to obtain the sensory evaluation of the inspection object non-destructively and easily without causing the loss of some inspection objects due to sampling testing or without having to secure skilled inspectors.

[0024] 2 is a block diagram showing an example of the functional configuration of an information processing device according to an embodiment. As shown in Fig. 2, the information processing device 1 includes a communication unit 10, an input unit 20, a display unit 30, a storage unit 40, and a control unit 50. As the information processing device 1, for example, a PC (Personal Computer) can be applied.

[0025] The communication unit 10 executes data communication with external devices, etc. via a network. For example, under the control of the control unit 50, the communication unit 10 acquires inspection image data 100 obtained by non-destructive inspection of an inspection object from an inspection device (not shown) and stores the image data in the storage unit 40.

[0026] For example, when an ultrasonic probe is used to probe an object to be inspected, data of an ultrasonic image obtained by the ultrasonic probe is acquired as inspection image data 100 from the inspection device.

[0027] In addition, under the control of the control unit 50, the communication unit 10 acquires a first training data set 41 used for machine learning of the first machine learning model 110 and a second training data set 42 used for machine learning of the second machine learning model 112 from an external device such as a data server and stores them in the memory unit 40.

[0028] The input unit 20 accepts operations from a user. The display unit 30 displays the processing results of the control unit 50. For example, the display unit 30 displays the evaluation result 103 inferred by the inference unit 53 of the control unit 50 on a display.

[0029] The storage unit 40 has a first training data set 41, a second training data set 42, autoencoder information 43, first model information 44, second model information 45, inspection image data 100, and evaluation results 103. For example, the storage unit 40 is realized by a memory or the like.

[0030] The first training data set 41 is a data set for each case that combines inspection image data of an inspection image obtained by non-destructive inspection of food with physicochemical analysis data showing the results of physicochemical analysis of the food.

[0031] The second training data set 42 is a data set for each case that combines physicochemical analysis data indicating the results of physicochemical analysis of foods with the evaluation results of sensory evaluation of the foods. Here, the foods in the first training data set and the foods in the second training data set are of the same type of food (e.g., fish such as tuna and yellowtail, and fruits such as watermelon and melon). However, the first training data set and the second training data set do not necessarily have to be of the same individual.

[0032] The autoencoder information 43 is information related to the autoencoder 111. The autoencoder information 43 includes, for example, the number of dimensions to be compressed. The autoencoder information 43 is set in advance by a user or the like.

[0033] The first model information 44 is information such as various parameters related to the first machine learning model 110. The setting values ​​of the parameters in the first model information 44 are updated by the machine learning of the first machine learning model 110 described above.

[0034] The second model information 45 is information such as various parameters related to the second machine learning model 112. The setting values ​​of the parameters in the second model information 45 are updated by the machine learning of the second machine learning model 112 described above.

[0035] The control unit 50 includes a first learning unit 51, a second learning unit 52, and an inference unit 53. For example, the control unit 50 is realized by a processor.

[0036] The first learning unit 51 is a processing unit that performs machine learning on the aforementioned first machine learning model 110. The second learning unit 52 is a processing unit that performs machine learning on the aforementioned second machine learning model 112. The inference unit 53 is a processing unit that infers the evaluation result 103 of the sensory evaluation using the first machine learning model 110 and the second machine learning model 112, based on the inspection image data 100 obtained by non-destructive testing of the inspection object.

[0037] Here, a detailed description will be given of the processes in the first learning unit 51, the second learning unit 52, and the inference unit 53. Fig. 3 is a flowchart showing an example of the operation of the information processing device according to the embodiment.

[0038] 3, when the process starts, the first learning unit 51 reads the first training data set 41 and performs learning of the autoencoder 111 (S1). FIG. 4 is an explanatory diagram illustrating learning of the autoencoder 111.

[0039] 4, the first learning unit 51 inputs the physicochemical analysis data 101a included in the first training data set 41 to the autoencoder 111, and trains the autoencoder 111 so that the data, which is dimensionally compressed and then restored to the original dimensions, becomes the physicochemical analysis data 101a. The first learning unit 51 generates an embedding (physicochemical information 102a) of the physicochemical analysis data 101a by dimensionality compression in this learning process.

[0040] 3, after S1, the first learning unit 51 uses the read first training data set 41 to learn the first machine learning model 110 (S2). FIG. 5 is an explanatory diagram illustrating the learning of the first machine learning model 110.

[0041] 5 , the first learning unit 51 performs machine learning on the first machine learning model 110 so that when test image data 100a included in the first training data set 41 is input to the first machine learning model 110, the output from the first machine learning model 110 becomes physicochemical information 102a generated from physicochemical analysis data 101a. Here, the first learning unit 51 stores the parameters of the first machine learning model 110 obtained by machine learning in the first model information 44.

[0042] 3, after S2, the second learning unit 52 reads the second training data set 42 and performs learning 112 (S3). FIG. 6 is an explanatory diagram illustrating learning of the second machine learning model 112.

[0043] 6 , the second learning unit 52 inputs the physicochemical analysis data 101b included in the second training data set 42 into the autoencoder 111 to perform dimensional compression, thereby generating embeddings (physicochemical information 102b) of the physicochemical analysis data 101b. Next, the second learning unit 52 performs machine learning on the second machine learning model 112 so that when the generated physicochemical information 102b is input to the second machine learning model 112, the output from the second machine learning model 112 becomes the evaluation result 103b included in the second training data set 42. Here, the second learning unit 52 stores the parameters of the second machine learning model 112 obtained by machine learning in second model information 45.

[0044] 3 , after S3, the inference unit 53 performs an inference process to infer an evaluation result 103 of the sensory evaluation based on the inspection image data 100, using the first machine learning model 110 and the second machine learning model 112. FIG. 7 is an explanatory diagram for explaining the inference of the evaluation result 103.

[0045] 7 , the inference unit 53 constructs a first machine learning model 110 and a second machine learning model 112 based on the first model information 44 and the second model information 45. Next, the inference unit 53 reads out the inspection image data 100 of the inspection target and inputs it into the constructed first machine learning model 110. As a result, the inference unit 53 acquires physicochemical information 102 corresponding to the inspection target (S4a).

[0046] Next, the inference unit 53 inputs the acquired physicochemical information 102 into the constructed second machine learning model 112. As a result, the inference unit 53 infers an evaluation result 103 of the sensory evaluation corresponding to the test object (S4b). The inference unit 53 displays the inferred evaluation result 103 on a display or the like via the display unit 30. At this time, the inference unit 53 may output the physicochemical information 102 from the display unit 30 together with the evaluation result 103 of the sensory evaluation corresponding to the test object.

[0047] As described above, the information processing device 1 inputs inspection image data 100 obtained by non-destructive testing of an inspection object into the first machine learning model 110, thereby acquiring physicochemical information 102 indicating the results of a physicochemical analysis of the inspection object. Here, the first machine learning model 110 is a model trained by machine learning using a first training dataset 41 that associates inspection images obtained by non-destructive testing of food with analytical information indicating the results of the physicochemical analysis of the food. The information processing device 1 inputs the acquired physicochemical information 102 into a second machine learning model 112, thereby inferring the evaluation result 103 of a sensory evaluation corresponding to the inspection object. Here, the second machine learning model 112 is a model trained by machine learning using a second training dataset 42 that associates analytical information indicating the results of the physicochemical analysis of the food with the results of the sensory evaluation of the food.

[0048] As a result, the information processing device 1 can nondestructively and easily obtain sensory evaluations of test objects without the need for sampling inspections, which can result in the loss of some test objects, or the need for skilled inspectors. As described above, in this embodiment, the first training data set and the second training data set do not necessarily contain identical food items. In conventional sampling inspections, inspectors taste the test objects, resulting in the loss of test objects (unavailable for shipment). The sensory evaluation described in this embodiment does not require the inspection of the test objects themselves, and can be achieved through nondestructive testing, thereby avoiding the loss of test objects. Furthermore, conventional sampling inspections are strictly limited to sensory evaluations of items that will never actually be shipped. In contrast, the sensory evaluation described in this embodiment can be performed on shippable (non-defective) items. The sensory evaluation method described in this embodiment is also applicable when the food items in the first training data set and the second training data set are identical, or when some of the items in each training data set are identical.

[0049] The first machine learning model 110 of the information processing device 1 is trained by machine learning using physicochemical information 102a obtained by dimensionally compressing food physicochemical analysis data 101 using an autoencoder 111. By training using physicochemical information 102 obtained by dimensionally compressing the physicochemical analysis data 101 in this manner, the information processing device 1 can prevent overlearning of the first machine learning model 110, which would otherwise be too well-suited to the food physicochemical analysis data 101.

[0050] In the information processing device 1, by applying an ultrasound image obtained by ultrasound exploration to the inspection image data 100, it is possible to obtain a sensory evaluation of the inspection object non-destructively and simply by ultrasound exploration of the inspection object.

[0051] The information processing device 1 outputs the physicochemical information 102 together with the evaluation result 103 of the sensory evaluation. This allows the user of the information processing device 1 to verify the relationship between the inspection image data 100 of the inspection target and the evaluation result 103 of the sensory evaluation through the physicochemical information 102. In other words, the interpretability of the evaluation result 103 of the sensory evaluation is improved.

[0052] The information processing device 1 trains a first machine learning model 110 by machine learning using a first training data set 41 that associates inspection images obtained by non-destructive inspection of foods with analytical information indicating the results of physicochemical analysis of the foods. The information processing device 1 trains a second machine learning model 112 by machine learning using a second training data set 42 that associates analytical information indicating the results of physicochemical analysis of foods with the results of sensory evaluation of the foods. This allows the information processing device 1 to obtain the first machine learning model 110 and the second machine learning model 112 for inferring the evaluation result 103 of the sensory evaluation corresponding to the inspection target from inspection image data 100 obtained by non-destructive inspection of the inspection target via physicochemical information 102.

[0053] Note that the components of each device shown in the figure do not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc.

[0054] For example, the training of the autoencoder 111 (S1), the training of the first machine learning model 110 (S2) and the training of the second machine learning model 112 (S3), and the inference processing (S4) may be performed by separate information processing devices.

[0055] Furthermore, the various processing functions of the first learning unit 51, the second learning unit 52, and the inference unit 53 performed by the control unit 50 of the information processing device 1 may be executed in whole or in part on a CPU (or a microcomputer such as an MPU or MCU (Micro Controller Unit)). Needless to say, the various processing functions may be executed in whole or in part on a program analyzed and executed by the CPU (or a microcomputer such as an MPU or MCU), or on hardware using wired logic. The various processing functions performed by the information processing device 1 may be executed by multiple computers working together using cloud computing.

[0056] The various processes described in the above embodiments can be realized by executing a program prepared in advance on a computer. Therefore, an example of a computer configuration (hardware) that executes a program having the same functions as those of the above embodiments will be described below. Fig. 8 is an explanatory diagram illustrating an example of the computer configuration.

[0057] 8, computer 200 has a CPU 201 that executes various types of arithmetic processing, an input device 202 that accepts data input, a monitor 203, and a speaker 204. Computer 200 also has a medium reading device 205 that reads programs and the like from a storage medium, an interface device 206 for connecting with various devices, and a communication device 207 for connecting with external devices via wired or wireless communication. Computer 200 also has a RAM 208 that temporarily stores various types of information, and a hard disk drive 209. Each unit (201 to 209) within computer 200 is connected to a bus 210.

[0058] The hard disk drive 209 stores a program 211 for executing various processes in the functional configuration (e.g., the first learning unit 51, the second learning unit 52, and the inference unit 53) described in the above embodiment. The hard disk drive 209 also stores various data 212 referenced by the program 211. The input device 202, for example, accepts input of operation information from an operator. The monitor 203, for example, displays various screens operated by the operator. The interface device 206 is connected to, for example, a printing device. The communication device 207 is connected to a communication network such as a LAN (Local Area Network) and exchanges various information with external devices via the communication network.

[0059] The CPU 201 reads out the program 211 stored in the hard disk drive 209, expands it in the RAM 208, and executes it to perform various processes related to the above-described functional configuration (e.g., the first learning unit 51, the second learning unit 52, and the inference unit 53). The program 211 does not have to be stored in the hard disk drive 209. For example, the program 211 stored in a storage medium readable by the computer 200 may be read out and executed. Examples of storage media readable by the computer 200 include portable storage media such as CD-ROMs, DVDs, and USB (Universal Serial Bus) memory, semiconductor memories such as flash memory, and hard disk drives. The program 211 may also be stored in a device connected to a public line, the Internet, a LAN, or the like, and the computer 200 may read out and execute the program 211 from the device.

[0060] REFERENCE SIGNS LIST 1... Information processing device 10... Communication unit 20... Input unit 30... Display unit 40... Memory unit 41... First training data set 42... Second training data set 43... Autoencoder information 44... First model information 45... Second model information 50... Control unit 51... First learning unit 52... Second learning unit 53... Inference unit 100, 100a... Inspection image data 101, 101a, 101b... Physicochemical analysis data 102, 102a, 102b... Physicochemical information 103, 103b... Evaluation results 110... First machine learning model 111... Autoencoder 112... Second machine learning model 200... Computer 201... CPU 202... Input device 203... Monitor 204... Speaker 205... Medium reading device 206... Interface device 207: Communication device 208: RAM 209: Hard disk device 210: Bus 211: Program 212: Various data

Claims

1. An inference processing program that causes a computer to execute the following processes: inputting an inspection image obtained by non-destructive testing of a food to be inspected into a first machine learning model trained by machine learning using a first training data dataset that matches an inspection image obtained by non-destructive testing of the food with first analytical information that indicates the results of a physicochemical analysis of the food, thereby obtaining target analytical information that indicates the results of the physicochemical analysis of the food; and inputting the obtained target analytical information into a second machine learning model trained by machine learning using a second training data dataset that matches second analytical information that indicates the results of the physicochemical analysis of the food with the results of a sensory evaluation of the food, thereby inferring a sensory evaluation result corresponding to the food to be inspected.

2. The inference processing program described in claim 1, characterized in that the first machine learning model is trained by the machine learning using information obtained by dimensionally compressing the first analytical information by an autoencoder.

3. The inference processing program according to claim 1, characterized in that the inspection image is an ultrasound image obtained by ultrasound exploration.

4. The inference processing program according to claim 1, characterized in that the inference processing outputs the target analysis information together with the sensory evaluation result.

5. A machine learning program that causes a computer to execute processes: training a first machine learning model that outputs target analytical information indicating physicochemical analysis results corresponding to a test target in response to input of an inspection image obtained by non-destructive testing of a first food, through machine learning using a first training data dataset that matches an inspection image obtained by non-destructive testing of the first food with first analytical information indicating the results of a physicochemical analysis of the first food; and training a second machine learning model that outputs a sensory evaluation result corresponding to a test target in response to input of target analytical information of the test target, through machine learning using a second training data dataset that matches second analytical information indicating the results of a physicochemical analysis of a second food with the results of a sensory evaluation of the second food.

6. The machine learning program described in claim 5, characterized in that the process of training the first machine learning model trains the first machine learning model by the machine learning using information obtained by dimensionally compressing the first analytical information by an autoencoder.

7. An inference processing method characterized by being executed by a computer to perform the following processes: inputting an inspection image obtained by non-destructive testing of an object to be inspected into a first machine learning model trained by machine learning using a first training data dataset that matches an inspection image obtained by non-destructive testing of a first food with first analytical information that indicates the results of a physicochemical analysis of the first food, thereby acquiring object analytical information that indicates the results of the physicochemical analysis of the object to be inspected; and inputting the acquired object analytical information into a second machine learning model trained by machine learning using a second training data dataset that matches second analytical information that indicates the results of a physicochemical analysis of a second food with the results of a sensory evaluation of the second food, thereby inferring the results of the sensory evaluation of the object to be inspected.

8. The inference processing method described in claim 7, characterized in that the first machine learning model is trained by the machine learning using information obtained by dimensionally compressing the first analytical information by an autoencoder.

9. The inference processing method according to claim 7, characterized in that the inspection image is an ultrasound image obtained by ultrasound exploration.

10. The inference processing method according to claim 7, characterized in that the inference processing outputs the target analysis information together with the sensory evaluation result.

11. A machine learning method characterized in that a computer executes the following processes: training a first machine learning model by machine learning using a first training data dataset that matches an inspection image obtained by non-destructive testing of a first food with first analytical information that indicates a physicochemical analysis result of the first food, in response to an input of an inspection image obtained by non-destructive testing of the inspection target, and outputting target analytical information that indicates a physicochemical analysis result corresponding to the inspection target; and training a second machine learning model by machine learning using a second training data dataset that matches second analytical information that indicates a physicochemical analysis result of a second food with a sensory evaluation result for the second food, in response to an input of target analytical information of the inspection target, and outputting a sensory evaluation result corresponding to the inspection target.

12. The machine learning method described in claim 11, characterized in that the process of training the first machine learning model trains the first machine learning model by the machine learning using information obtained by dimensionally compressing the first analytical information by an autoencoder.

13. An information processing device comprising a control unit that executes processing to obtain object analytical information indicating the physicochemical analysis results corresponding to the object of inspection by inputting an inspection image obtained by non-destructive testing of an object of inspection into a first machine learning model trained by machine learning using a first training data dataset that matches an inspection image obtained by non-destructive testing of a first food with first analytical information indicating the physicochemical analysis results of the first food, and to infer the sensory evaluation results corresponding to the object of inspection by inputting the obtained object analytical information into a second machine learning model trained by machine learning using a second training data dataset that matches second analytical information indicating the physicochemical analysis results of a second food with the sensory evaluation results of the second food.

14. The information processing device according to claim 13, characterized in that the first machine learning model is trained by the machine learning using information obtained by dimensionally compressing the first analytical information by an autoencoder.

15. The information processing device according to claim 13, wherein the inspection image is an ultrasound image obtained by ultrasound exploration.

16. The information processing device according to claim 13, characterized in that the inference process outputs the target analysis information together with the sensory evaluation result.

17. An information processing device comprising a control unit that executes processing to train a first machine learning model by machine learning using a first training data dataset that matches an inspection image obtained by non-destructive testing of a first food with first analytical information that indicates the results of a physicochemical analysis of the first food, in response to an input of an inspection image obtained by non-destructive testing of the inspection object, and to train a second machine learning model by machine learning using a second training data dataset that matches second analytical information that indicates the results of a physicochemical analysis of a second food with the results of a sensory evaluation of the second food, in response to an input of the object analytical information of the inspection object, and to train a second machine learning model by machine learning using a second training data dataset that matches second analytical information that indicates the results of a physicochemical analysis of the second food with the results of a sensory evaluation of the second food, in response to an input of the object analytical information of the inspection object, and 18. The information processing device described in claim 17, characterized in that the process of training the first machine learning model trains the first machine learning model by the machine learning using information obtained by dimensionally compressing the first analytical information by an autoencoder.

Images