Identification apparatus, identification method, generation method, generation apparatus, program, and recording medium

Abstract

To provide a technique related to identification of a growth state of an animal included in an image.SOLUTION: An information processing apparatus (1) includes an acquisition unit (11) that acquires a target image, an identification unit (13) that identifies a position on the target image, distinction between life and death, and a degree of growth of at least any of a target animal and a trace of the target animal included in the target image by applying identification processing using a learned detection model to the target image, and an output data generation unit (14) that generates output data including an identification result.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an identification device, an identification method, a generation method, a generation device, a program, and a recording medium for identifying a living thing included in an image. [Background technology]

[0002] Techniques for detecting animals contained in images are known.

[0003] For example, Non-Patent Document 1 discloses a model incorporating an algorithm that detects whitefly eggs from an image of whitefly eggs attached to a leaf and counts the number of eggs. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Micha Gracianna Devi1, Dan Jeric Arcega Rustia, Lize Braat, Kas Swinkels, Federico Fornaguera Espinosa, Bart M. van Marrewijk, Jochen Hemming and Lotte Caarls1,"Eggsplorer: a rapid plant-insect resistance determination tool using an automated whitefly egg quantification algorithm", May 20, 2023 Summary of the Invention [Problem to be solved by the invention]

[0005] As an example, technology for detecting animals contained in an image is used in the development of pesticides against pests such as whiteflies. In the development of such pesticides, technology for identifying the growth state of animals such as whiteflies is required for efficient development. For example, technology for identifying the age of whiteflies contained in an image, whether they are alive or dead, whether there are signs of emergence, etc. is required. However, although the model described in the above-mentioned Non-Patent Document 1 can detect whitefly eggs as an example of an animal, it has a problem in that it is difficult to identify the growth state of the animal.

[0006] An object of one aspect of the present invention is to provide a technique for identifying the growth state of an animal included in an image. [Means for solving the problem]

[0007] In order to solve the above problem, an identification device according to one embodiment of the present invention includes an acquisition unit that acquires a target image, an identification unit that applies an identification process to the target image using a trained detection model to identify the position on the target image, whether it is alive or dead, and the degree of growth of at least one of a target animal and a trace of the target animal contained in the target image, and an output data generation unit that generates output data including the identification result by the identification unit.

[0008] In order to solve the above problem, an identification method according to one aspect of the present invention includes an acquisition step of acquiring a target image, an identification step of applying an identification process using a trained detection model to the target image to identify the position on the target image, whether it is alive or dead, and the degree of growth of at least one of a target animal and a trace of the target animal contained in the target image, and an output data generation step of generating output data including the identification result obtained by the identification step.

[0009] In order to solve the above problem, a generation method according to one aspect of the present invention includes an acquisition step of acquiring training data including a training image and correct labels for at least one of a target animal and a trace of the target animal contained in the training image, the correct labels for the position on the training image, whether it is alive or dead, and the degree of growth, and a detection model generation step of generating, by referring to the training data, a detection model that detects the position on the target image, whether it is alive or dead, and the degree of growth for at least one of a target animal and a trace of the target animal contained in the target image.

[0010] In order to solve the above problem, a generation device according to one aspect of the present invention includes an acquisition unit that acquires training data including a training image and correct labels for at least one of a target animal and a trace of the target animal contained in the training image, where the correct labels are for the position on the training image, whether alive or dead, and the degree of growth, and a detection model generation unit that generates, by referring to the training data, a detection model that detects the position on the target image, whether alive or dead, and the degree of growth for at least one of a target animal and a trace of the target animal contained in the target image.

[0011] The identification device according to each aspect of the present invention may be realized by a computer. In this case, the identification device program that causes the computer to operate as each part (software element) of the identification device to realize the identification device on a computer, and the computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention.

[0012] The generating device according to each aspect of the present invention may be realized by a computer. In this case, the generating device program that causes the computer to operate as each part (software element) of the generating device to realize the generating device on a computer, and the computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention. [Effects of the Invention]

[0013] According to one aspect of the present invention, a technique for identifying the growth state of an animal included in an image can be provided. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a block diagram showing a configuration of an information processing system according to a first embodiment of the present invention. [Figure 2] 3 is a flowchart showing an example of a flow of processing executed by the information processing device according to the first embodiment of the present invention. [Figure 3] FIG. 4 is a diagram illustrating an example of processing executed by a division unit according to the first embodiment of the present invention. [Figure 4] FIG. 2 is a diagram illustrating an example of a configuration of a detection model according to the first embodiment of the present invention. [Figure 5] FIG. 4 is a diagram illustrating an example of processing executed by an update unit according to the first embodiment of the present invention. [Figure 6] FIG. 10 is a flowchart showing another example of the flow of the process executed by the information processing device according to the first embodiment of the present invention. [Figure 7] FIG. 3 is a diagram illustrating an example of processing executed by a recognition unit according to the first embodiment of the present invention. [Figure 8] FIG. 2 is a diagram illustrating an example of a user interface according to the first embodiment of the present invention. [Figure 9] FIG. 2 is a diagram showing how a user inputs data into a user interface according to the first embodiment of the present invention. [Figure 10] FIG. 4 is a diagram showing an example of an analysis result output in the first embodiment of the present invention. [Figure 11] FIG. 10 is a block diagram showing the configuration of an information processing system according to a second embodiment of the present invention. [Figure 12] 1 is a graph showing the results of Examples 1 and 2 of the present invention. [Figure 13] 1 is a graph showing the results of Example 3 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0015] [Embodiment 1] Hereinafter, one embodiment of the present invention will be described in detail.

[0016] (Information processing system 100) The configuration of the information processing system 100 will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the information processing system 100 according to this embodiment. As shown in Fig. 1, the information processing system 100 includes an information processing device 1, an imaging device 110, and a terminal device 120. An example of the information processing device 1 is a server. An example of the imaging device 110 is a camera. An example of the terminal device 120 is a smartphone, a tablet, or a PC (Personal Computer).

[0017] The information processing device 1, the imaging device 110, and the terminal device 120 are connected to each other via a network so that they can communicate with each other, as shown in Fig. 1. The specific configuration of the network is not particularly limited, but examples that can be used include a wireless LAN (Local Area Network), a wired LAN, a WAN (Wide Area Network), a public line network, a mobile data communication network, or a combination of these networks.

[0018] In the information processing system 100, the imaging device 110 captures a target image TI including at least one of a target animal and a trace of the target animal. The imaging device 110 outputs the captured target image TI to the information processing device 1.

[0019] The information processing device 1 is an identification device that identifies the position, whether alive or dead, and the degree of growth of at least one of a target animal and a trace of the target animal included in the target image TI on the target image TI by applying an identification process using a trained detection model DM. The detection model DM is a model generated to detect the position, whether alive or dead, and the degree of growth of at least one of a target animal and a trace of the target animal included in the target image TI on the target image. Note that in this embodiment, the "distinguishing between alive and dead" and "identifying the degree of growth" of the target animal are examples of "identifying the growth state" of the target animal.

[0020] Furthermore, the information processing device 1 presents the output data OUT including the classification result to the user, or alternatively, the information processing device 1 may output the output data OUT including the classification result to the terminal device 120.

[0021] The terminal device 120 presents the identification result included in the output data OUT to the user of the terminal device 120.

[0022] The target image TI is not particularly limited as long as it includes a target animal, but as an example in this embodiment, we will explain the case where the target image TI is an image that includes a target animal and / or a plant with traces of the target animal attached.

[0023] Furthermore, the target animal is not particularly limited, but as an example in this embodiment, the target animal is an arthropod, and the information processing system 100 will be described as identifying the degree of growth by identifying at least one of whether it is an egg, the age, and the emergence trace.

[0024] Furthermore, the type of arthropod is not limited, but in this embodiment, as an example, the case where the arthropod is an insect or a mite will be described. Furthermore, the type of insect or mite is not particularly limited, but as an example, the insect or mite is a whitefly, scale insect, spider mite, planthopper, aphid, thrips, or leafhopper. In this embodiment, as an example, the case where the target animal is a whitefly will be described. That is, in this embodiment, the detection model DM is a model that detects at least one of the following as the degree of growth of a whitefly: whether it is an egg, the instar, and the emergence trace.

[0025] Furthermore, in the information processing system 100, the information processing device 1 is also a generation device that generates (trains) a detection model DM by referring to training data LD including training images LI and supervised labels GL. The training data LD includes training images LI and supervised labels GL relating to at least one of a target animal and a trace of the target animal included in the training images LI, which are supervised labels GL relating to a position on the training images LI, a distinction between alive and dead, and a growth rate.

[0026] (Information processing device 1) The configuration of the information processing device 1 will be described again with reference to Fig. 1. As shown in Fig. 1, the information processing device 1 includes a control unit 10, a storage unit 20, a communication unit 30, and an input / output unit 40.

[0027] (Storage unit 20) The storage unit 20 stores data referenced by the control unit 10. Examples of the storage unit 20 include, but are not limited to, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0028] 1, examples of data stored in the storage unit 20 include learning data LD, target image TI, detection model DM, detection result DR, analysis result AR, and output data OUT. The learning data LD, target image TI, detection model DM, and output data OUT are as described above.

[0029] 1, the training images LD include a training image group consisting of a plurality of training images LI and a correct label group consisting of a plurality of correct labels GL. The training images LI and the correct labels LG are associated with each other and stored in the storage unit 20.

[0030] The detection result DR is data indicating the result of detection by the detection model DM. The analysis result AR is data obtained by the information processing device 1 analyzing the classification result obtained in the classification process.

[0031] (Communication unit 30) The communication unit 30 is an interface for transmitting and receiving data via a network. Examples of the communication unit 30 include, but are not limited to, communication chips for various communication standards such as Ethernet (registered trademark), Wi-Fi (registered trademark), and wireless communication standards for mobile data communication networks, and USB-compliant connectors.

[0032] (Input / output section 40) The input / output unit 40 is an interface with an input device that accepts input of data and an output device that outputs data. Examples of input devices include, but are not limited to, a microphone, a camera, an eye-gaze input device, a keyboard, and a touchpad. Examples of output devices include, but are not limited to, a speaker and a liquid crystal display.

[0033] (Control unit 10) The control unit 10 controls each component included in the information processing device 1. The control unit 10 also includes an acquisition unit 11, a learning unit 12, a classification unit 13, and an output data generation unit 14, as shown in FIG.

[0034] (Acquisition part 11) The acquisition unit 11 acquires data via the communication unit 30 or the input / output unit 40. As one example, the acquisition unit 11 acquires a target image TI output from the imaging device 110. As another example, the acquisition unit 11 acquires learning data LD. As yet another example, the acquisition unit 11 acquires a user instruction output from the input / output unit 40 or the terminal device 120. The acquisition unit 11 stores the acquired data in the storage unit 20.

[0035] (Study Section 12) The learning unit 12 generates a learning model. As an example, the learning unit 12 (detection model generation unit) generates a detection model DM with reference to the learning data LD. The learning unit 12 stores the generated detection model DM in the storage unit 20. Furthermore, the learning unit 12 learns the detection model DM with reference to the learning data LD.

[0036] As shown in FIG. 1, the learning unit 12 includes a dividing unit 121 and an updating unit 122.

[0037] The dividing unit 121 divides an image into a plurality of images. As an example, the dividing unit 121 divides a learning image LI into a plurality of partial images PI. The dividing unit 121 supplies the plurality of partial images PI to the updating unit 122.

[0038] The update unit 122 updates the parameters of the learning model. As an example, the update unit 122 performs an update process to update one or more parameters of the detection model DM so that the difference between the detection result DR by the detection model DM to which the learning image LI has been input and the correct label GL becomes smaller. Furthermore, the update unit 122 inputs each of the multiple partial images PI generated by the division unit 121 to the detection model DM, thereby performing the update process for each of the multiple partial images PI.

[0039] An example of the processing executed by the dividing unit 121 and the updating unit 122 will be described later.

[0040] (Identification unit 13) The classification unit 13 classifies the growth state of the target animal included in the target image TI. As an example, the classification unit 13 applies a classification process using the trained detection model DM to the target image TI to classify the position on the target image TI, whether it is alive or dead, and the degree of growth of at least one of the target animal and the target animal's trace included in the target image TI. Furthermore, the classification unit 13 classifies at least one of whether it is an egg, the age, and the emergence trace as the degree of growth.

[0041] As shown in FIG. 1, the identification unit 13 includes a division unit 131, an integration unit 132, and an analysis unit 133.

[0042] The dividing unit 131 divides the target image TI into a plurality of partial images PI. When the dividing unit 131 divides the target image TI into a plurality of partial images PI, the classification unit 13 inputs each of the plurality of partial images PI to a detection model DM, thereby performing the classification process on each of the plurality of partial images PI.

[0043] The integration unit 132 integrates the detection results DR. As an example, the integration unit 132 executes integration processing to integrate the detection results DR of the plurality of partial images PI. The integration unit 132 supplies the integrated detection result DR to the analysis unit 133 as the identification result.

[0044] The analysis unit 133 analyzes the classification result obtained by the classification unit 13. The analysis unit 133 supplies the analysis result to the output data generation unit .

[0045] Examples of the processes executed by the dividing unit 131, the integrating unit 132, and the analyzing unit 133 will be described later.

[0046] (Output data generation unit 14) The output data generation unit 14 generates output data OUT to be output via the communication unit 30 or the input / output unit 40. As an example, the output data generation unit 14 generates output data OUT including the classification results by the classification unit 13. The output data generation unit 14 also generates output data OUT including the classification results integrated by the integration unit 132. The output data generation unit 14 also generates output data OUT including the analysis results analyzed by the analysis unit 133. Examples of the output data OUT will be described later.

[0047] (Example 1 of the flow of processing executed by the information processing device 1) An example of the flow of the process (generation method) executed by the information processing device 1 will be described with reference to Fig. 2. Fig. 2 is a flow diagram showing an example of the flow of the process executed by the information processing device 1 according to this embodiment.

[0048] (Step S11: Acquisition step) In step S11, the acquisition unit 11 acquires training data LD including a training image group that is a plurality of training images LI and a correct label group that is a plurality of correct labels GL related to at least one of a target animal and a trace of the target animal included in the training images LI, the position on the training image LI, whether it is alive or dead, and the degree of growth. The acquisition unit 11 stores the acquired training image group and correct label group in the storage unit 20.

[0049] (Step S12: Detection model generation step) In step S12, the learning unit 12 generates, from the target image TI, a detection model DM that detects the position on the target image TI, whether it is alive or dead, and the degree of growth of at least one of the target animal and the trace of the target animal contained in the target image TI, by referring to the learning data LD. Details of the processing in the detection model generation step will be described below.

[0050] (Step S121) In step S121, the division unit 121 divides the learning image LI into a plurality of partial images PI. The division unit 121 also inputs the plurality of partial images PI to the detection model DM. Here, the correct label GL acquired in step S11 may be the correct label GL of the partial image PI.

[0051] An example of the process executed by the dividing unit 121 in step S121 will be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of the process executed by the dividing unit 121 according to this embodiment.

[0052] The dividing unit 121 first generates a cropped learning image C_LI by cropping at least a portion of a region C_R from the learning image LI. For example, the dividing unit 121 sets at least a portion of a region C_R from the learning image LI, which is 34 mm high and 51 mm wide, as shown in the upper part of FIG. 3. Then, as shown in the center of FIG. 3, the dividing unit 121 generates a cropped learning image C_LI by cropping out a region C_R that is 23.8 mm high and 35.6 mm wide. The cropped region is not particularly limited, but an example is a rectangular region whose size is 70% of the size of the learning image LI. With this configuration, the dividing unit 121 can crop out an image of an in-focus region from the learning image LI.

[0053] Next, the dividing unit 121 divides the cut-out learning image C_LI into a plurality of partial images PI. For example, the dividing unit 121 divides the cut-out learning image C_LI into a plurality of partial images PI each having a height of 2.77 mm and a width of 3.69 mm, as shown in the lower part of FIG.

[0054] The dividing unit 121 may also use data augmentation. For example, the dividing unit 121 may generate a partial image PI by randomly applying processing to the partial image PI, such as changing the color, enlarging or reducing the image, cutting out a part of the image, or inverting the image.

[0055] (Step S122) In step S122, the update unit 122 executes an update process to update one or more parameters of the detection model DM so as to reduce the difference between the detection result DR by the detection model DM to which the learning image LI has been input and the correct label GL. More specifically, the update unit 122 executes the update process for each of the multiple partial images PI by inputting each of the multiple partial images PI to the detection model DM.

[0056] An example of the configuration of the detection model DM will be described with reference to Fig. 4. Fig. 4 is a diagram showing an example of the configuration of the detection model DM according to this embodiment.

[0057] The specific configuration of the detection model DM is not limited, but as an example, as shown in FIG. 4, the detection model DM may be a model that uses an SSD (Single Shot Multibox Detector).

[0058] In this case, as shown in Figure 4, the detection model DM inputs the input target image TI to the base network BN used for image classification and the convolutional layer CNN. An example of a network used as the base network BN is VGG-16. The base network BN and the convolutional layer CNN generate features for boxes of different sizes (e.g., 38x38 boxes, 19x19 boxes, etc.). The detection model DM then outputs the detection result DR by applying a non-maximum suppression algorithm to the generated features.

[0059] The update unit 122 updates the parameters by referring to the feature amounts of the boxes of different sizes output from the base network BN and the convolutional layer CNN. An example of the processing executed by the update unit 122 will be described with reference to Fig. 5. Fig. 5 is a diagram showing an example of the processing executed by the update unit 122 according to this embodiment.

[0060] 5, the update unit 122 refers to the position coordinates of the 5x5 BOX candidates, the detection result DR by the detection model DM, and the position of the whitefly on the training image LI indicated by the correct label GL included in the training data LD, and calculates the error using the following formula (1). The position coordinates of the BOX candidates may be generated in advance with various sizes and aspect ratios. Error = offset_t - offset_e (1) Here, offset_t indicates the difference (Offset1 to Offset4 shown in FIG. 5) between the position of the whitefly on the training image LI indicated by the correct label GL included in the training data LD and one or more position coordinates of the BOX candidates. Also, offset_e indicates the difference between the position of the whitefly detected by the detection model DM and one or more position coordinates of the BOX candidates.

[0061] The update unit 122 updates one or more parameters of the detection model DM so as to reduce the calculated error. The update unit 122 also refers to feature amounts other than those of the 5×5 box in the same way, and performs update processing so as to reduce the error.

[0062] As another example, the update unit 122 refers to a probability distribution according to the detection model DM as shown in Fig. 5. Specifically, the update unit 122 calculates the cross entropy error H using the following equation (2).

[0063]

number

[0064] Here, p(x) is the probability distribution given by the learning data LD, and q(x) is the probability distribution according to the detection model DM. The update unit 122 executes an update process to update one or more parameters of the detection model DM so as to reduce the calculated cross-entropy error H.

[0065] (Step S123) In step S123, the learning unit 12 determines whether or not a convergence condition is satisfied. As an example, the learning unit 12 determines whether or not the error calculated in step S122 has converged within a predetermined range.

[0066] If it is determined in step S123 that the convergence condition is not satisfied (step S123: NO), the dividing unit 121 executes step S121 again.

[0067] Here, the dividing unit 121 may change the partial images PI to be input to the detection model DM every time the parameter update process is executed (mini-batch learning). For example, in step S121, the dividing unit 121 selects a predetermined number of partial images PI (e.g., 30). The dividing unit 121 inputs the selected predetermined number of partial images PI to the detection model DM.

[0068] Next, if it is determined in step S123 that the convergence condition is not satisfied and step S121 is executed again, the division unit 121 reselects a predetermined number of partial images PI that are different in at least some respects from the predetermined number of partial images PI selected previously, and inputs the reselected predetermined number of partial images PI to the detection model DM.

[0069] (Step S124) If it is determined in step S123 that the convergence condition is satisfied (step S123: YES), the learning unit 12 outputs the learned detection model DM. As an example, the learning unit 12 stores the learned detection model DM in the memory unit 20. As another example, the learning unit 12 outputs the learned detection model DM to the terminal device 120 via the communication unit 30.

[0070] In this way, the detection model DM is generated by the generation method shown in FIG.

[0071] (Example 2 of flow of processing executed by information processing device 1) Another example of the flow of the process (identification method) executed by the information processing device 1 will be described with reference to Fig. 6. Fig. 6 is a flow diagram showing another example of the flow of the process executed by the information processing device 1 according to this embodiment.

[0072] (Step S21: Acquisition step) In step S21, the acquisition unit 11 acquires a target image TI. The acquisition unit 11 stores the acquired target image TI in the storage unit 20.

[0073] (Step S13: Identification step) In step S13, the identification unit 13 applies an identification process using the trained detection model DM to the target image TI, thereby identifying the position on the target image TI, whether it is alive or dead, and the degree of growth of at least one of the target animal and the trace of the target animal contained in the target image TI. Details of the process in the identification step S13 will be described.

[0074] (Step S131) In step S131, the dividing unit 131 divides the target image TI into a plurality of partial images PI. Then, the classification unit 13 inputs each of the plurality of partial images PI into a detection model DM.

[0075] An example of the process executed by the dividing unit 131 in step S131 will be described with reference to Fig. 7. Fig. 7 is a diagram showing an example of the process executed by the identifying unit 13 according to this embodiment.

[0076] The dividing unit 131 first generates a cut-out identification image C_TI by cutting out at least a part of the region C_R from the target image TI, similar to the above-described dividing unit 121. For example, as shown in the upper left of Fig. 7, the dividing unit 131 generates a cut-out identification image C_TI by cutting out a region C_R measuring 23.8 mm in height and 35.6 mm in width.

[0077] Next, the dividing unit 131 divides the cut-out identification image C_TI into a plurality of partial images PI, as shown on the right side of Fig. 7. As an example, the dividing unit 131 divides the cut-out identification image C_TI into 691 x 517. Here, the dividing unit 131 divides the image into a plurality of partial images PI having overlapping regions where the image overlaps with each other.

[0078] Then, the classification unit 13 inputs each of the plurality of partial images PI into the detection model DM.

[0079] (Step S132) In step S132, the integration unit 132 acquires the detection results DR for each of the plurality of partial images PI from the detection model DM, and then performs integration processing to integrate the detection results DR obtained by the detection model DM.

[0080] An example of the process executed by the integration unit 132 in step S132 will be described with reference again to FIG.

[0081] The integration unit 132 determines whether or not a target animal has been detected in the overlapping region. For example, as shown in the center left of Fig. 7, a case will be described in which a whitefly has been detected in an overlapping region where partial images PI_1 and PI_2 overlap each other.

[0082] When a whitefly is detected in an overlapping region, the integrating unit 132 determines whether the detected whitefly is the same whitefly. As an example, a case will be described in which the detection model DM outputs one or more detection results DR (the detection result DR for the partial image PI_1 and the detection result DR for the partial image PI_2) for each of multiple partial images PI (partial image PI_1 and partial image PI_2), each including the reliability of detection. The reliability output by the detection model DM includes the reliability of the detection position (bounding box) and the reliability of the identification (whether it is alive or dead, how many instars it is, and whether it is an emergence mark or an egg).

[0083] In this case, if the reliability of each of the detection results DR for the partial images PI_1 and PI_2 is equal to or greater than a predetermined value, the integrating unit 132 determines that the whitefly detected in the overlapping region of the partial images PI_1 and PI_2 is the same whitefly. Then, the integrating unit 132 removes the detection of the whitefly in the overlapping region of either the partial image PI_1 or the partial image PI_2. That is, the integrating process performed by the integrating unit 132 includes a process of removing overlapping detections from the detection results DR by the detection model DM in the overlapping region by referring to the reliability.

[0084] The integration unit 132 executes integration processing to generate an integrated detection result DR as shown in the lower left of Fig. 7. Then, the integration unit 132 supplies the integrated detection result DR to the analysis unit 133 as an identification result.

[0085] (Step S133) In step S133, the analysis unit 133 analyzes the identification results by the identification unit 13. As an example, the analysis unit 133 counts the number of live target animals, the number of dead target animals, the number of target animals by age, the number of emergence marks, and the number of eggs. Then, the analysis unit 133 supplies the analysis results including the identification results to the output data generation unit 14.

[0086] As another example, the analysis unit 133 analyzes the mortality rate, egg-laying suppression, growth inhibition, and emergence inhibition. As one example, the analysis unit 133 calculates the mortality rate based on the number of living target animals and the number of dead target animals.

[0087] As another example, the analysis unit 133 calculates the difference between the number of eggs analyzed from the identification results before a certain pesticide is sprayed and the number of eggs analyzed from the identification results after the pesticide is sprayed. If the calculated difference is greater than a predetermined value, the analysis unit 133 determines that the pesticide is effective in suppressing egg laying. Similarly, the analysis unit 133 analyzes growth inhibition based on age and emergence inhibition based on emergence scars.

[0088] (Step S14: Output data generation step) In step S14, the output data generation unit 14 generates output data OUT including the classification result by the classification unit 13 in step S13. In other words, the output data generation unit 14 generates output data OUT including the detection result DR by the detection model DM. The output data generation unit 14 may also generate output data OUT including the analysis result in addition to the classification result.

[0089] (Example of user interface) As an example, the control unit 10 of the information processing device 1 outputs the output data OUT generated by the output data generation unit 14 to the input / output unit 40. The input / output unit 40 presents the output data OUT to the user via a connected output device. Furthermore, the control unit 10 may receive, via the input / output unit 40, a user instruction regarding the output data OUT presented to the user.

[0090] An example of a user interface UI for receiving user instructions will be described with reference to Figs. 8 to 10. Fig. 8 is a diagram showing an example of the user interface UI according to this embodiment. Fig. 9 is a diagram showing how a user inputs information in the user interface UI according to this embodiment. Fig. 10 is a diagram showing an example of an analysis result output in this embodiment.

[0091] The control unit 10 causes the output device to display the user interface UI including the images and UI elements included in the output data OUT. In other words, the images and UI elements included in the user interface UI are included in the output data OUT.

[0092] As an example, the control unit 10 causes the output device to display a user interface UI shown in Fig. 8. As shown in Fig. 8, the user interface UI includes an interface FI that receives an input to select a target image TI to be analyzed, and an interface ASI that receives an input to start the analysis. When the control unit 10 receives an input to the interface ASI, it displays the target image TI selected in the interface FI and the classification result for the target image TI.

[0093] Furthermore, the target image TI included in the user interface UI may be a detection result image that visually shows the detection result DR by the detection model DM, as shown in Fig. 8. For example, as shown on the right side of Fig. 8, an index IR1 indicating whether the target animal detected in the target image TI is an egg, a first instar, a second instar, a third instar, a fourth instar, an emergence mark, or a dead insect, and a reliability IR2 may be associated with the target animal detected in the target image TI. Furthermore, a cause of death may be associated with the target animal detected in the target image TI. This configuration can be realized by having the information processing device 1 train the detection model DM to also identify the cause of death.

[0094] As shown in FIG. 9, the user interface UI also includes a selection item SI for selecting at least one of the objects included in the detection result DR, and a correction start interface MSI and a correction reflection interface MRI, which are second UI elements for accepting user instructions regarding the selected object.

[0095] For example, the control unit 10 receives an input to select the interface MSI, and then receives an input to select an object on the object image TI. In this case, the control unit 10 displays a selection item SI that allows the user to select whether the received object is an egg, first instar, second instar, third instar, fourth instar, emergence trace, or dead insect. Furthermore, the control unit 10 receives an input to select the correction-reflecting interface MRI, after receiving an input for the selection item SI.

[0096] The acquisition unit 11 acquires information indicating an input for the selection item SI and an input to select the correction reflecting interface MRI via the input / output unit 40. In other words, the acquisition unit 11 acquires a user instruction via the correction reflecting interface MRI. The acquisition unit 11 supplies the acquired user instruction to the learning unit 12.

[0097] The update unit 122 of the learning unit 12 updates one or more parameters of the detection model DM with reference to the user's instruction. For example, if the user's instruction is for a target animal whose detection result of the detection model DM is "first stage," the learning unit 12 updates one or more parameters of the detection model DM so that the detection result of the target animal becomes "second stage."

[0098] 9, the user interface UI includes an analysis result AR. Also, as shown in FIG. 9, the user interface UI may include an interface DOI that accepts an input to output the analysis result AR. When the control unit 10 accepts an input to select the interface DOI, it outputs the analysis result AR. As an example, as shown in FIG. 10, the control unit 10 outputs an analysis result AR that counts the number of live target animals, the number of dead target animals, the number of target animals by age, the number of eclosion marks, and the number of eggs.

[0099] As described above, the information processing device 1 may output the output data OUT to the terminal device 120. In this case, the terminal device 120 presents a user interface UI to the user and outputs instructions from the user to the information processing device 1.

[0100] (Effects of information processing device 1) In this way, the information processing device 1 applies a classification process using the detection model DM to the target image TI, thereby identifying the position on the target image TI, whether it is alive or dead, and the degree of growth of at least one of the target animal and the trace of the target animal included in the target image TI. Therefore, the information processing device 1 can identify the growth state of the animal included in the image.

[0101] [Embodiment 2] Other embodiments of the present invention will be described below. For ease of explanation, the same reference numerals will be used to designate components having the same functions as those described in the above embodiment, and the description thereof will not be repeated.

[0102] The configuration of the information processing system 200 will be described with reference to Fig. 11. Fig. 11 is a block diagram showing the configuration of the information processing system 200 according to this embodiment. As shown in Fig. 11, the information processing system 200 includes an identification device 2, an imaging device 110, and a terminal device 120. The imaging device 110 and the terminal device 120 are as described above.

[0103] In the information processing system 200, the classification device 2 classifies the position, live / dead status, and growth level of at least one of a target animal and a trace of the target animal included in the target image TI captured by the imaging device 110. The classification device 2 also presents output data OUT including the classification result to the user. That is, the classification device 2 is a device having the functions of the classification device provided in the information processing device 2 described above.

[0104] (Identification device 2) 11, the identification device 2 includes a control unit 10A, a storage unit 20A, a communication unit 30, and an input / output unit 40. The communication unit 30 and the input / output unit 40 are as described above. An example of the processing executed by the identification device 2 is also the same as that described with reference to FIG. 8.

[0105] (Storage unit 20A) The storage unit 20A stores data referenced by the control unit 10A. Examples of the storage unit 20A include, but are not limited to, a flash memory, an HDD, an SSD, or a combination thereof.

[0106] Examples of data stored in the storage unit 20A include a target image TI, a detection model DM, a detection result DR, an analysis result AR, and output data OUT, as shown in Fig. 11. The target image TI, the detection model DM, the detection result DR, the analysis result AR, and the output data OUT are as described above.

[0107] (Control unit 10A) The control unit 10A controls each component included in the classification device 2. As shown in Fig. 11, the control unit 10A also includes an acquisition unit 11, a classification unit 13, and an output data generation unit 14. The acquisition unit 11, the classification unit 13, and the output data generation unit 14 are as described above.

[0108] (Effect of Identification Device 2) In this way, the classification device 2 applies a classification process using the detection model DM to the target image TI, thereby identifying the position on the target image TI, whether it is alive or dead, and the degree of growth of at least one of the target animal and the trace of the target animal included in the target image TI. Therefore, the classification device 2 can also identify the growth state of the animal included in the image.

[0109] Example 1 Test data was input into the detection model DM generated under the following conditions and evaluation was performed. Only one whitefly is visible in the test image - Only one type of label is learned (1st to 4th instars, all emergence marks are learned as whiteflies) The training data was divided into 691x517, with 3000 partial images (random samples). 477 test data The evaluation was performed using a PR curve with two indices. Specifically, the vertical axis is Precision and the horizontal axis is Recall, and the values ​​where Precision = Recall were used as the representative precision. Precision = (number of positive detections) / (number of positive detections + number of false detections) Recall = (number of positive detections) / (number of positive detections + number of negative detections) In addition, for correct detection and incorrect detection, (A∩B) / (A∪B)>0.5 was defined as a correct detection between the correct position (bounding box, hereinafter referred to as "A") and the detected position (bounding box, hereinafter referred to as "B").

[0110] The graph for this case is shown in the upper part of Fig. 12. Fig. 12 is a graph showing the results of Examples 1 and 2. As shown in the upper part of Fig. 12, the representative accuracy was 96%, and the detection performance of the detection model DM was extremely high.

[0111] Example 2 Test data was input to the detection model DM generated under the following conditions and evaluation was performed. The evaluation method was the same as that described above. -The number of whiteflies in the test image is multiple. - Only one type of label is learned (1st to 4th instars, all emergence marks are learned as whiteflies) The training data was divided into 691x517, with 3000 partial images (random samples). 309 test data The graph for this case is shown in the lower part of Fig. 12. As shown in the lower part of Fig. 12, the representative accuracy was 93%, and the detection performance of the detection model DM was extremely high.

[0112] Example 3 A test image was input to the detection model DM generated under the following conditions, and an evaluation was performed. The evaluation method was the same as that described above. -The number of whiteflies in the test image is multiple. - There are 7 types of labels to learn (1st to 4th instars, emergence marks, dead insects, eggs) The training data was divided into 691x517, with 3000 partial images (random samples). 701 test data The graph for this case is shown in Figure 13. Figure 13 is a graph showing the results of Example 3. As shown in Figure 13, the representative accuracy for each was as follows: 1st instar: 93% 2nd instar: 87% 3rd instar: 87% 4th instar: 96% Emergence marks: 94% Eggs: 51% Dead insects: 71% As such, the detection performance of the detection model DM was very high for first to fourth instars and emergence marks. The detection model DM was able to detect eggs and dead insects. Furthermore, the accuracy of the detection model for eggs and dead insects can be improved by training it using more data.

[0113] According to the above configuration, it is possible to identify the growth state of animals in images, which allows for the efficient development of pesticides for pests such as whiteflies. Such effects also contribute to the achievement of Goal 2 of the United Nations' Sustainable Development Goals (SDGs), such as "End hunger, achieve food security and improved nutrition, and promote sustainable agriculture."

[0114] [Software implementation example] The functions of the information processing device 1 and the identification device 2 (hereinafter referred to as "devices") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 10, 10A).

[0115] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program, thereby realizing the functions described in each of the above embodiments.

[0116] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0117] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0118] Furthermore, each process described in each of the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI ​​may run on the control device or on another device (for example, an edge computer or a cloud server).

[0119] [summary] The identification device according to aspect 1 of this embodiment includes an acquisition unit that acquires a target image, an identification unit that applies an identification process to the target image using a trained detection model to identify the position on the target image, whether the target animal is alive or dead, and the degree of growth of at least one of the target animal and traces of the target animal contained in the target image, and an output data generation unit that generates output data including the identification result by the identification unit.

[0120] With the above configuration, the classification device can classify the growth state of an animal included in an image.

[0121] In the identification device of aspect 2 of this embodiment, in aspect 1 above, the target animal is an arthropod, and the identification unit identifies at least one of whether the animal is an egg, its age, and emergence marks as the degree of growth.

[0122] With the above configuration, the identification device can identify whether an arthropod contained in an image is an egg, its instar, and whether it is an emergence mark.

[0123] In the classification device according to aspect 3 of this embodiment, the classification unit in aspect 1 or 2 divides the target image into a plurality of partial images, and inputs each of the plurality of partial images into the detection model, thereby performing the classification process for each of the plurality of partial images.

[0124] With the above configuration, the classification device can classify the growth state of an animal even if the animal included in the image is small.

[0125] In the classification device according to aspect 4 of this embodiment, the classification unit in aspect 3 performs an integration process to integrate the detection results of each of the multiple partial images, and the output data generation unit generates the output data including the integrated detection results.

[0126] With the above configuration, even when the classification device performs classification processing on each of the partial images obtained by dividing an image, it can output a classification result using the image before division.

[0127] In the identification device according to aspect 5 of this embodiment, the detection model according to aspect 4 outputs one or more detection results for each of the plurality of partial images, including the reliability of each detection, and the plurality of partial images include a plurality of partial images having overlapping regions that overlap each other, and the integration process by the identification unit includes a process of removing overlapping detections from the detection results by the detection model in the overlapping regions by referring to the reliability.

[0128] With the above configuration, the classification device can output a suitable classification result even when the image is divided into partial images having overlapping regions.

[0129] In the identification device according to aspect 6 of this embodiment, the target image in any of aspects 1 to 5 above is an image including a plant having at least one of the target animal and a trace of the target animal attached thereto.

[0130] With the above configuration, the identification device can identify the growth state of an animal attached to a plant.

[0131] In the identification device according to Aspect 7 of this embodiment, the arthropod in any of Aspects 2 to 6 above is an insect or a mite.

[0132] With the above configuration, the identification device can identify the growth state of insects or mites contained in an image.

[0133] In the identification device according to aspect 8 of this embodiment, the insects or mites in aspect 7 above are whiteflies, scale insects, spider mites, planthoppers, aphids, thrips or leafhoppers.

[0134] With the above configuration, the identification device can identify whiteflies, scale insects, spider mites, planthoppers, aphids, thrips, or leafhoppers contained in an image.

[0135] In the classification device according to aspect 9 of this embodiment, the output data generated by the output data generation unit in aspects 1 to 8 above includes a classification result image that visually shows the classification result by the classification unit, a first UI element for selecting at least one of the objects included in the classification result, and a second UI element for receiving user instructions regarding the selected object.

[0136] With the above configuration, the identification device can present the identification result and further receive a user's instruction regarding the object included in the identification result.

[0137] In the identification device according to Aspect 10 of this embodiment, the output data generation section in Aspect 9 generates output data including an analysis result obtained by analyzing the identification result by the identification section.

[0138] With the above configuration, the identification device can present an analysis result obtained by analyzing the identification result.

[0139] In the identification device according to aspect 11 of this embodiment, the acquisition unit in aspects 1 to 10 above further acquires training data including a training image and correct labels for at least one of the target animal and the target animal's traces contained in the training image, including correct labels for the position on the training image, whether alive or dead, and the degree of growth, and the identification device further includes a learning unit that refers to the training data to train a detection model.

[0140] With the above configuration, the identification device can learn a detection model.

[0141] In the identification device according to aspect 12 of this embodiment, the learning unit in aspect 11 above updates one or more parameters of the detection model so as to reduce the difference between the detection result obtained by the detection model to which the learning image is input and the correct label.

[0142] With the above configuration, the identification device can preferably learn the detection model.

[0143] The identification method according to aspect 13 of this embodiment includes an acquisition step of acquiring a target image, an identification step of applying an identification process using a trained detection model to the target image to identify the position on the target image, whether the target animal is alive or dead, and the degree of growth of at least one of the target animal and traces of the target animal contained in the target image, and an output data generation step of generating output data including the identification result obtained by the identification step.

[0144] With the above configuration, the identification method can identify the growth state of an animal included in an image.

[0145] The generation method according to aspect 14 of this embodiment includes an acquisition step of acquiring training data including a training image and correct labels for at least one of a target animal and a trace of the target animal contained in the training image, the correct labels for the position on the training image, whether alive or dead, and the degree of growth, and a detection model generation step of generating, by referring to the training data, a detection model that detects the position on the target image, whether alive or dead, and the degree of growth for at least one of a target animal and a trace of the target animal contained in the target image.

[0146] With the above configuration, the generation method can generate a detection model that detects the growth state of an animal included in an image.

[0147] In the generation method according to aspect 15 of this embodiment, the target animal in aspect 14 above is an arthropod, and the detection model is a model that detects at least one of the following as the degree of growth: whether it is an egg, its age, and the presence of eclosion marks.

[0148] With the above configuration, the generation method can generate a detection model that detects whether an arthropod included in an image is an egg, an instar, and an emergence trace.

[0149] In the generation method according to aspect 16 of this embodiment, in the detection model generation step in aspect 14 or 15 above, an update process is performed to update one or more parameters of the detection model so that the difference between the detection result by the detection model to which the learning image is input and the correct label is reduced.

[0150] With the above configuration, the generation method can generate a suitable detection model.

[0151] In the generation method according to aspect 17 of this embodiment, in the detection model generation step in any of aspects 14 to 16 above, the learning image is divided into a plurality of partial images, and each of the plurality of partial images is input into the detection model, thereby performing the update process for each of the plurality of partial images.

[0152] With the above configuration, the generation method can generate a detection model that can detect the growth state of an animal even if the animal included in the image is small.

[0153] In the generating method according to Aspect 18 of this embodiment, the target image in any of Aspects 14 to 17 above is an image including a plant having at least one of the target animal and a trace of the target animal attached thereto.

[0154] With the above configuration, the generation method can generate a detection model that detects the growth state of an animal attached to a plant.

[0155] In the production method according to Aspect 19 of this embodiment, the arthropod in any one of Aspects 13 to 18 above is an insect or a mite.

[0156] With the above configuration, the generation method can generate a detection model that detects the growth state of insects or mites included in an image.

[0157] In the production method according to Aspect 20 of this embodiment, the insects or mites in Aspect 19 above are whiteflies, scale insects, spider mites, planthoppers, aphids, thrips or leafhoppers.

[0158] With the above configuration, the generation method can generate a detection model that can detect whiteflies, scale insects, spider mites, planthoppers, aphids, thrips, or leafhoppers contained in an image.

[0159] The detection model according to aspect 21 of this embodiment is a detection model generated by the generation method according to any one of aspects 13 to 20 above.

[0160] With the above configuration, the detection model can suitably detect the growth state of an animal included in an image.

[0161] The generation device according to aspect 22 of this embodiment includes an acquisition unit that acquires training data including a training image and correct labels for at least one of a target animal and a trace of the target animal contained in the training image, where the correct labels are for the position on the training image, whether alive or dead, and the degree of growth, and a detection model generation unit that generates, by referring to the training data, a detection model that detects the position on the target image, whether alive or dead, and the degree of growth for at least one of a target animal and a trace of the target animal contained in the target image.

[0162] With the above configuration, the generating device can generate a detection model that can detect the growth state of an animal included in an image.

[0163] The generation device according to aspect 23 of this embodiment, in aspect 22 above, further includes an output data generation unit that generates output data including detection results obtained by the detection model, and the output data generated by the output data generation unit includes a detection result image that visually shows the detection results obtained by the detection model, a first UI element for selecting at least one of the objects included in the detection results, and a second UI element for receiving user instructions regarding the selected object.

[0164] With the above configuration, the generating device can present the classification result and further receive a user's instruction regarding the target included in the classification result.

[0165] In the generation device relating to aspect 24 of this embodiment, the acquisition unit in aspect 23 acquires the user's instructions via the second UI element, and the detection model generation unit updates one or more parameters of the detection model by referring to the user's instructions.

[0166] With the above configuration, the generating device can generate a detection model with updated parameters based on a user instruction.

[0167] The program according to aspect 25 of this embodiment is a program for causing a computer to function as the identification device described in any one of aspects 1 to 12 above, and causes the computer to function as the acquisition unit, the identification unit, and the output data generation unit.

[0168] With the above configuration, the program can cause the computer to function as an identification device that identifies the growth state of an animal included in an image.

[0169] A recording medium according to aspect 26 of this embodiment is a computer-readable recording medium having the program according to aspect 25 recorded thereon.

[0170] With the above configuration, the recording medium can record a program that causes a computer to function as an identification device that identifies the growth state of an animal included in an image.

[0171] A program according to aspect 27 of this embodiment is a program for causing a computer to function as the generating device according to any one of aspects 22 to 24 above, and causes the computer to function as the acquisition unit and the detection model generating unit.

[0172] With the above configuration, the program can cause the computer to function as a generating device that generates a detection model that detects the growth state of an animal included in an image.

[0173] A recording medium according to aspect 28 of this embodiment is a computer-readable recording medium having the program according to aspect 27 recorded thereon.

[0174] With the above configuration, the recording medium can record a program that causes a computer to function as a generating device that generates a detection model that detects the growth state of an animal included in an image.

[0175] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0176] 1. Information processing equipment 2. Identification device 11 Acquisition Department 12 Learning Department 13 Identification unit 14 Output data generation unit 121, 131 division part 122 Update Department 132 Integration Department 133 Analysis Department AR analysis results DM detection model DR detection results GL correct label LD learning data LI Learning Images OUT Output data PI Partial Image TI target image

Claims

1. an acquisition unit that acquires a target image; By applying a discrimination process using a trained detection model to the target image, at least one of a target animal and a trace of the target animal contained in the target image is detected. a position on the target image; the distinction between life and death, and Degree of growth an identification unit for identifying the an output data generation unit that generates output data including the classification result by the classification unit; An identification device comprising:

2. the target animal is an arthropod, The identification unit indicates the degree of growth. Egg or not, Age, and Emergence marks Identify at least one of The identification device according to claim 1 .

3. The identification unit Dividing the target image into a plurality of partial images; The classification process is performed for each of the plurality of partial images by inputting each of the plurality of partial images into the detection model. The identification device according to claim 2 .

4. the identification unit performs an integration process to integrate detection results of the plurality of partial images; The output data generation unit generates the output data including the integrated detection result. The identification device according to claim 3 .

5. the detection model outputs one or more detection results for each of the plurality of partial images, each detection result including a reliability of the detection; the plurality of partial images include a plurality of partial images having overlapping regions that overlap each other; The integration process by the identification unit includes: and removing overlapping detections from the detection results by the detection model in the overlapping region by referring to the reliability. The identification device according to claim 4 .

6. The target image is an image including a plant having at least one of the target animal and a trace of the target animal attached thereto. The identification device according to any one of claims 1 to 5.

7. The arthropod is an insect or a mite The identification device according to any one of claims 2 to 5.

8. The insects or mites are whiteflies, scale insects, spider mites, planthoppers, aphids, thrips or leafhoppers. The identification device according to claim 7.

9. The output data generated by the output data generation unit includes: a classification result image visually showing the classification result by the classification unit; and a first UI element for selecting at least one object included in the identification result; a second UI element for receiving a user instruction regarding the selected object; 6. The identification device according to claim 1, further comprising:

10. The output data generation unit generates output data including an analysis result obtained by analyzing the classification result by the classification unit. The identification device according to claim 9.

11. the acquisition unit acquires a learning image and a correct answer label relating to at least one of a target animal and a trace of the target animal included in the learning image, a position on the learning image; the distinction between life and death, and Degree of growth Further obtain learning data including the correct answer label for The identification device further includes a learning unit that refers to the learning data and learns a detection model. The identification device according to any one of claims 1 to 5.

12. The learning unit One or more parameters of the detection model are updated so that the difference between the detection result by the detection model to which the learning image is input and the correct label becomes small. The identification device according to claim 11.

13. an acquisition step of acquiring a target image; By applying a discrimination process using a trained detection model to the target image, at least one of a target animal and a trace of the target animal contained in the target image is detected. a position on the target image; the distinction between life and death, and Degree of growth an identification step of identifying an output data generating step of generating output data including the identification result obtained by the identifying step; An identification method including:

14. A learning image and a correct answer label for at least one of a target animal and a trace of the target animal included in the learning image, a position on the learning image; the distinction between life and death, and Degree of growth An acquisition step of acquiring learning data including a correct answer label for From the target image, information relating to at least one of the target animal and the trace of the target animal contained in the target image is obtained. a position on the target image; the distinction between life and death, and Degree of growth a detection model generation step of generating a detection model for detecting the A generating method including:

15. the target animal is an arthropod, The detection model is Egg or not, Age, and Emergence marks It is a model that detects at least one of The method of claim 14.

16. In the detection model generation step, An update process is performed to update one or more parameters of the detection model so that the difference between the detection result by the detection model to which the learning image is input and the correct label becomes small. The method of claim 15.

17. In the detection model generation step, Dividing the learning image into a plurality of partial images; The update process is performed for each of the plurality of partial images by inputting each of the plurality of partial images into the detection model.

17. The method of claim 16.

18. The target image is an image including a plant having at least one of the target animal and a trace of the target animal attached thereto.

18. A method according to any one of claims 14 to 17.

19. The arthropod is an insect or a mite 18. A method according to any one of claims 15 to 17.

20. The insects or mites are whiteflies, scale insects, spider mites, planthoppers, aphids, thrips or leafhoppers.

20. The method of claim 19.

21. The detection model generated by the generation method according to any one of claims 14 to 17.

22. A learning image and a correct answer label for at least one of a target animal and a trace of the target animal included in the learning image, a position on the learning image; the distinction between life and death, and Degree of growth an acquisition unit that acquires learning data including a correct answer label for From the target image, information relating to at least one of the target animal and the trace of the target animal contained in the target image is obtained. a position on the target image; the distinction between life and death, and Degree of growth a detection model generation unit that generates a detection model for detecting the A generating device comprising:

23. an output data generation unit that generates output data including a detection result by the detection model; The output data generated by the output data generation unit includes: a detection result image visually showing a detection result by the detection model; and a first UI element for selecting at least one target included in the detection result; a second UI element for receiving a user instruction regarding the selected object; 23. The generating device of claim 22, comprising:

24. the acquisition unit acquires an instruction from the user via the second UI element; The detection model generation unit updates one or more parameters of the detection model by referring to an instruction from the user.

24. The generating device of claim 23.

25. A program for causing a computer to function as the identification device according to claim 1, the program causing a computer to function as the acquisition unit, the identification unit, and the output data generation unit.

26. A computer-readable recording medium on which the program according to claim 25 is recorded.

27. A program for causing a computer to function as the generating device according to claim 22, the program causing a computer to function as the acquisition unit and the detection model generating unit.

28. A computer-readable recording medium on which the program according to claim 27 is recorded.

Images