Insect recognition model generator, and insect production management system
The insect recognition model generation device automates cricket counting by using machine learning to recognize and segment insects, reducing manual labor and improving efficiency in cricket inventory management.
Patent Information
- Application Number
- JP2023079200
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-12
- Publication Date
- 2024-11-22
AI Technical Summary
Existing methods for counting the number of crickets require manual labor, which is inefficient and labor-intensive.
An insect recognition model generation device equipped with a machine learning device and storage device that generates an insect recognition model through training and test image data, using a camera to acquire images, input units for annotation, and correction units to refine the model for accurate insect recognition.
Automates the counting process, reducing the manual effort required to count insects by generating a model that accurately identifies and segments insects, thereby decreasing the number of man-hours needed for inventory.
Smart Images

Figure 2024163512000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an insect recognition model generating device and an insect production management system. [Background technology]
[0002] Patent Document 1 describes a breeding device for breeding crickets, a type of insect. It describes that the breeding device provides a container in which absorbent cotton soaked in water is placed in the breeding container to provide water to the crickets, and provides a large number of lightly rolled up newspapers in the breeding container so that the crickets can hide in the shade. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] Japanese Patent Application Publication No. 10-191834 Summary of the Invention [Problem to be solved by the invention]
[0004] According to the conventional technology, adult crickets are counted and fed into a shipping cardboard box at the time of shipping. In other words, in the conventional technology, the number of crickets is counted manually. Therefore, there is a need to reduce the labor required for counting crickets.
[0005] The present invention has been made in consideration of the above problems, and aims to provide an insect recognition model generation device and an insect production management system that can reduce the labor required for counting the number of insects. [Means for solving the problem]
[0006] One aspect of the present invention is An insect recognition model generation device for generating an insect recognition model for recognizing insects, comprising: A machine learning device and a storage device, The storage device includes: Training image data for learning in which the insects are linked by annotations; Test image data for evaluation; A trained model generated by performing machine learning using the training image data and the test image data, The machine learning device includes: an image acquisition unit that acquires the training image data and the test image data; an input unit to which the annotation to be assigned to the training image data is input; A model generation unit that generates the trained model by performing machine learning using the training image data; a determination unit that determines whether the insect included in the test image data is recognized when the test image data is determined by the trained model; The insect recognition model generation device includes an annotation correction unit that corrects the annotation if the judgment result of the test image data by the judgment unit is incorrect, and maintains the annotation if the judgment result is correct, based on the judgment result of the test image data by the judgment unit.
[0007] Another aspect of the present invention is An insect production management system comprising any one of the insect recognition model generation devices described above. Effect of the Invention
[0008] According to one aspect of the present invention, an insect recognition model capable of recognizing insects can be generated. This allows the machine learning device to count the number of segmentations recognized as insects, thereby allowing the number of insects to be calculated. As a result, the task of manually counting the number of insects can be omitted, thereby reducing the man-hours required for counting insects.
[0009] According to another aspect of the present invention, the labor required for counting crickets in an insect production management system can be reduced. [Brief description of the drawings]
[0010] [Figure 1] 1 is a block diagram showing an insect production management system and an insect recognition model generation device of embodiment 1. [Diagram 2] FIG. 1 is a cross-sectional view showing a breeding box according to a first embodiment. [Diagram 3] A diagram to explain about crickets. [Figure 4] FIG. 1 shows a state in which annotations according to embodiment 1 have been added to a cricket. [Diagram 5] 4 is a flowchart showing the operation of the insect recognition model generating device of the first embodiment. [Figure 6] FIG. 2 is a diagram showing training image data according to the first embodiment. [Figure 7] FIG. 13 is a diagram showing a state in which annotations have been added to training image data in the first embodiment. [Figure 8] FIG. 1 shows a state in which a cricket has been segmented by a trained model according to embodiment 1. [Figure 9] FIG. 4 is a diagram showing test image data according to the first embodiment. [Figure 10] FIG. 4 is a view showing an example of a determination result of test image data according to the first embodiment. [Figure 11] A figure showing the state in which a corrected annotation has been added to a cricket in embodiment 1. [Figure 12] 1A and 1B are diagrams showing trained models in accordance with embodiment 1, where (a) is a diagram showing a previous trained model, and (b) is a diagram showing a newly generated trained model. [Figure 13] FIG. 13 is a diagram showing the state in which annotations have been added to a two-spotted cricket in accordance with embodiment 2. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0011] (Embodiment 1) The present invention relates to an insect recognition model generating device 1 and an insect production management system 2 including the same. The insects according to this embodiment are not particularly limited, and may be ones that undergo complete metamorphosis or ones that undergo incomplete metamorphosis when they grow. Insects that undergo incomplete metamorphosis are insects whose larvae and adults have similar shapes and that metamorphose from larvae to adults by repeatedly molting without going through a pupal period. For example, Orthoptera can be suitably used as insects that undergo incomplete metamorphosis. Examples of Orthoptera include crickets, locusts, and grasshoppers. In this embodiment, a cricket C1 is exemplified. In the following description, some of the configurations of multiple identical members may be given reference symbols, and the reference symbols may be omitted for other configurations.
[0012] 1. Overview of Insect Production Management System 2 1, the insect production management system 2 includes a rearing device 10 and an insect recognition model generation device 1. Crickets C1 are reared in the rearing device 10. The insect recognition model generation device 1 includes a machine learning device 20 and a storage device 30.
[0013] The machine learning device 20 includes a camera 21 (an example of an image acquisition unit), an input unit 22, a model generation unit 23, a determination unit 24, and an annotation correction unit 25. The storage device 30 stores training image data 31, test image data 32, and a trained model 33 (an example of an insect recognition model). The storage device 30 is not particularly limited, and any storage medium such as a semiconductor memory or a hard disk drive can be appropriately selected.
[0014] 2. Rearing apparatus 10 and cricket C1 As shown in Fig. 2, the rearing device 10 is used for rearing crickets C1. Although not shown in detail, the rearing device 10 has a plurality of rearing boxes 11 connected by connecting members 12. The rearing boxes 11 are formed in a box shape. A plurality of crickets C1 are housed in each rearing box 11. Each rearing box 11 houses a water supplying means (not shown) for providing water to the crickets C1 and a feeding means for providing food to the crickets C1.
[0015] Crickets C1 of the same age are reared in the rearing boxes 11. It is known that crickets C1 grow in multiple stages by increasing their size while repeating the so-called "molting" ecology of renewing the tissue on the surface of their bodies. Therefore, in this example, for the sake of convenience, one unit period determined by the ecology of molting is defined as "one instar," and crickets C1 are reared in each of the multiple rearing boxes 11 at the same age.
[0016] The age of the cricket C1 during the period from hatching to the first molting (first molting) can be defined as the "1st instar". In this case, the age of the cricket C1 during the period from the first molting to the next molting (second molting) is the "2nd instar", and the age of the cricket C1 during the period from the second molting to the next molting (third molting) is the "3rd instar". In this way, in an insect having a molting ecology, the age of the organism during the period from after the (M-1)th molt to the Mth molt can be defined as the "Mth instar" (where M is a natural number).
[0017] In this example, when the crickets C1 grow to the eighth instar, the rearing is terminated and the crickets C1 are collected, and eight rearing boxes 11 are prepared. Figure 2 shows the rearing boxes 11 in which the first instar crickets C1 are reared. However, the number of rearing boxes 11 is not limited to eight, and may be any number.
[0018] Two adjacent rearing boxes 11 are connected by a connecting member 12. The rearing boxes 11 are provided with a through hole 13 for attaching the connecting member 12. The connecting member 12 has a communication passage 12a connecting the two rearing boxes 11. Therefore, one rearing box 11 and the other rearing box 11 communicate with each other by the communication passage 12a of the connecting member 12.
[0019] The communication passage 12a of the connection member 12 not only connects the rearing boxes 11 to each other, but also has a cross-sectional area set to allow the movement of 1st instar crickets C1 and to prevent the movement of 2nd instar or older crickets C1. This is the same for the other connection members 12, although not shown. This allows the rearing boxes 11 to rear crickets C1 of the same age.
[0020] 2, a camera 21 is attached to the upper end of the side wall of the rearing box 11 via a stay 14. The camera 21 can capture image data of the crickets C1 reared in the rearing box 11. Although not shown in detail, the camera 21 is attached to each rearing box 11.
[0021] Now, the cricket C1 will be described with reference to Fig. 3. The cricket C1 in the first to sixth instars has a head 41, a thorax 42, and an abdomen 43. The head 41 has two eyes 41a. Two antennae 41b extend from the head 41. Six legs 42a extend from the thorax 42. Two tail hairs 43a protrude from the end of the abdomen 43. From the first to sixth instars, the size of the body increases with age. However, although the balance of the relative sizes of the head 41, thorax 42, and abdomen 43, and the legs 42a, changes, the morphology hardly changes.
[0022] When the female cricket C1 reaches the 7th to 8th instars, one ovipositor 43b extending from the abdomen 43 is formed. This makes it possible to distinguish between male and female crickets C1 from the 7th instar onwards.
[0023] As described above, the cricket C1 grows by repeatedly molting. The exoskeleton 44a of the male cricket C1 in the 1st to 6th instars and the 7th to 8th instars is almost the same as that of the cricket C1, except that it does not have the eye 41a. In addition, the exoskeleton 44a has no contents and has wrinkles 44c on its surface, which is different from the cricket C1 body.
[0024] The shell 44b of the female cricket C1 in the 7th to 8th instars has a shape corresponding to the ovipositor 43b. In this respect, it is the same as the shape of the female cricket C1 in the 7th to 8th instars, but it also differs from the main body of the cricket C1 in that it does not have the eye 41a and has wrinkles 44c on its surface.
[0025] When the cricket C1 becomes an adult, the wings 45 are formed. Tip portions 45a of the wings 45 protrude from the tip of the abdomen 43. In this way, the adult and larval stages of the cricket C1 can be distinguished from each other by the presence or absence of the wings 45.
[0026] 3. Insect recognition model generation device 1 The insect recognition model generating device 1 will be described with reference to Fig. 1 and Fig. 4. As shown in Fig. 1, a camera 21 captures image data of a cricket C1. The captured image data is stored in a storage device 30. Of the image data, data used as training data in machine learning is designated as training image data 31 for learning. Moreover, of the image data, data used as test data in machine learning is designated as test image data 32 for evaluation.
[0027] For the training image data 31, annotations 50 are added to the captured image of a cricket C1 via the input unit 22 (see FIG. 4). As the input unit 22, any input device such as a mouse, a touch panel, or a joystick can be appropriately selected.
[0028] As shown in FIG. 4, the operator uses the input unit 22 to surround the outer shapes of the head 41, thorax 42, and abdomen 43 of the cricket C1 captured in the training image data 31 with a closed line. The closed line is not particularly limited and may be a polygon such as a triangle, a rectangle, a pentagon, or a hexagon, or may be a circle or an ellipse. In this embodiment, the head 41, thorax 42, and abdomen 43 of the cricket C1 are surrounded by a hexagon. This hexagon identifies the body part of the cricket C1 and is set as the body annotation 50a. The cricket C1 captured in the training image data 31 is identified by this body annotation 50a. In addition, information about the shape of the cricket C1 is linked to the body annotation 50a by the six vertices of the hexagon and the number of pixels in the area surrounded by the hexagon. The information about the shape of the cricket C1 is not particularly limited, and examples thereof include body length, body width, and the like.
[0029] Further, for a female cricket C1 in the 7th to 8th instars, in addition to a hexagonal body annotation 50a surrounding the head 41, thorax 42, and abdomen 43, a rectangular ovipositor annotation 50b surrounding the ovipositor 43b is given.
[0030] Additionally, for both male and female adult insects, a triangular wing annotation 50c is further added surrounding the wing 45. Note that the shape of the wing annotation 50c is not limited to a triangle, and any shape such as a square or a pentagon may be appropriately adopted.
[0031] The annotation 50 is not added to the shells 44a and 44b. That is, the main body annotation 50a, the ovipositor annotation 50b, and the wing annotation 50c are not added to the shells 44a and 44b.
[0032] In the following description, when there is no need to distinguish between the main body annotation 50a, the ovipositor annotation 50b, and the wing annotation 50c, they may be referred to as the annotation 50. The types of annotation 50 are not limited to the main body annotation 50a, the ovipositor annotation 50b, and the wing annotation 50c.
[0033] 1, the model generation unit 23 generates a trained model 33 by performing machine learning using the training image data 31 to which the annotations 50 have been added via the input unit 22. The trained model 33 is stored in the storage device 30.
[0034] The judgment unit 24 acquires the test image data 32 and the trained model 33 stored in the memory unit, and judges the test image data 32 using the trained model 33.
[0035] The annotation correction unit 25 corrects the annotation 50 if it is incorrect based on the judgment result of the judgment unit 24 on the test image data 32, and maintains the annotation 50 if it is correct.
[0036] 4. Operation of the insect recognition model generation device 1 The operation of the insect recognition model generating device 1 will be described with reference to Fig. 5 to Fig. 12. Fig. 5 shows a main flow of the operation of the insect recognition model generating device 1. When the insect recognition model generating device 1 is started, the camera 21 captures an image of a cricket C1 being reared in the rearing device 10 as training image data 31 (S1). As shown in Fig. 6, a plurality of training image data 31 are captured.
[0037] Next, annotations 50 are added to the photographed training image data 31 (S2). As shown in Fig. 7, the worker adds a body annotation 50a by surrounding the outer shapes of the head 41, thorax 42, and abdomen 43 of the cricket C1 in the training image data 31 with a hexagonal figure. However, if the cricket C1 has an ovipositor 43b, an ovipositor annotation 50b may be added. Also, if the cricket C1 has wings 45, a wing annotation 50c may be added.
[0038] Next, returning to FIG. 5, a trained model 33 is generated using the training image data 31 to which the main annotation 50a has been added (S3). The trained model 33 is a knowledge model for recognizing a cricket C1 from among objects captured in the image data. As shown in FIG. 8, for example, when the object captured in the image data is recognized as a cricket C1, the periphery of the cricket C1 is surrounded by a segmentation 51a. However, the shape of the segmentation 51a is not limited, and may be a polygon such as a triangle, a square, a pentagon, or a hexagon, or any shape such as a circle or an oval can be appropriately selected, and in this embodiment, it is formed as a hexagon.
[0039] The segmentation 51a may include information corresponding to the body annotation 50a. The segmentation 51a may also include information corresponding to the ovipositor annotation 50b, or may include information corresponding to the wing annotation 50c.
[0040] Next, returning to Fig. 5, a plurality of test image data 32 are captured (S4) for evaluating the generated trained model 33. As shown in Fig. 9, a plurality of test image data 32 different from the training image data 31 are captured. However, the test image data 32 may be a part of a plurality of image data captured of a cricket C1 as the training image data 31, and the other part as the test image data 32.
[0041] Next, the judgment unit 24 judges the test image data 32 using the trained model 33. Fig. 10 shows an example of the judgment result of the test image data 32. Four crickets C1 are photographed in Fig. 10. Of the four crickets C1, three crickets C1 except for the cricket C1 at the bottom right are recognized as crickets C1 in the judgment using the trained model 33 and are surrounded by the segmentation 51a.
[0042] On the other hand, the cricket C1 located at the bottom right of FIG. 10 was not recognized as a cricket C1 when determined using the trained model 33, and therefore was not assigned a segmentation 51a.
[0043] Next, returning to FIG. 5, in a judgment using the trained model 33, it is determined whether or not the cricket C1 has been recognized (S6).
[0044] In FIG. 10, the three crickets C1 except for the cricket C1 at the bottom right are correctly determined to be crickets C1. Therefore, the determination result by the trained model 33 is correct. That is, the cricket C1 has been correctly recognized (S6: Y in FIG. 5). In this case, the annotation correction unit 25 maintains the main annotation 50a (S7 in FIG. 5). With the above, the operation of the insect recognition model generation device 1 is completed.
[0045] In FIG. 10, the cricket C1 at the bottom right is not recognized as the cricket C1. That is, the cricket C1 is not correctly recognized (S6:N in FIG. 5). In this case, the worker gives a new body annotation 50a to the cricket C1 that could not be correctly recognized, as shown in FIG. 11. As a result, the annotation correction unit 25 corrects the body annotation 50a (S8 in FIG. 5). Returning to S3 in FIG. 5, a new trained model 33 is generated based on the corrected body annotation 50a. As shown in FIG. 12, a new trained model 33 is generated for the cricket C1 that could not be correctly recognized by the previous trained model 33 based on the corrected body annotation 50a, and a new segmentation 51b is given. This improves the recognition accuracy of the cricket C1.
[0046] 5. Effects of this Form Next, the effects of this embodiment will be described. According to this embodiment, it is possible to generate a trained model 33 capable of recognizing the cricket C1. The machine learning device 20 counts the number of segmentations 51a and 51b recognized as the cricket C1, and thereby the number of crickets C1 can be calculated. This makes it possible to omit the task of manually counting the number of crickets C1, thereby reducing the man-hours required for counting the crickets C1.
[0047] According to this embodiment, the body annotation 50a is made by surrounding the outline of the cricket C1 with a closed line. The figure formed by this closed line is a polygon having multiple vertices, and the body annotation 50a includes information on the relative positions of the multiple vertices of the polygon and information based on the number of pixels of the polygon. This makes it possible to obtain information based on the relative position coordinates of the multiple vertices of the polygon and the number of pixels equivalent to the area of the polygon. This makes it possible to improve the accuracy of insect recognition.
[0048] In addition, the cricket C1, which is an insect that undergoes incomplete metamorphosis, has a similar shape as a larva and an adult, so it is difficult to distinguish between the larva and the adult with the naked eye. According to this embodiment, the larva and the adult of the cricket C1, which is an insect that undergoes incomplete metamorphosis, can be distinguished by the trained model 33. This improves the recognition accuracy of the cricket C1.
[0049] In addition, the cricket C1, which is an insect that undergoes incomplete metamorphosis, grows by molting, and thus produces many exoskeletons 44a, 44b. The shapes of these exoskeletons 44a, 44b are similar to the shape of the cricket C1, making it difficult to distinguish them with the naked eye. According to this embodiment, the exoskeletons 44a, 44b of the cricket C1 can be distinguished from insects by the trained model 33. This improves the recognition accuracy of the cricket C1.
[0050] Moreover, the cricket C1, which is an insect that undergoes incomplete metamorphosis, grows as the larvae repeatedly molt. Each time the larvae molt, the age increases. For this reason, the shape of the incomplete metamorphosis cricket C1 does not change much before and after molting. For this reason, it is difficult to distinguish between different ages of larvae with the naked eye. According to this embodiment, the ages of the larvae of the cricket C1 can be distinguished by the trained model 33.
[0051] Furthermore, according to this embodiment, by adding the wing annotation 50c, it is possible to distinguish whether the cricket C1, which is an incomplete metamorphosis insect, is an adult or a larva, depending on whether or not it has wings 45.
[0052] According to this embodiment, by adding the ovipositor annotation 50b, it is possible to distinguish between male and female crickets C1 based on the presence or absence of an ovipositor 43b.
[0053] The insect production control system 2 according to this embodiment includes an insect recognition model generation device 1. This makes it possible to reduce the number of steps required for counting the crickets C1 in the insect production control system 2.
[0054] (Embodiment 2) Next, a second embodiment will be described with reference to Fig. 13. In this embodiment, a two-spotted cricket C2 (an example of an insect) is bred. When the two-spotted cricket C2 becomes an adult, two patterns 46 spaced apart from each other appear on the back side of the thorax 42. This makes it possible to distinguish between the larvae and adults of the two-spotted cricket C2.
[0055] In this embodiment, instead of the wing annotation 50c, the outline of the thorax 42 is surrounded by a closed line in correspondence with the pattern 46 of the two-spotted cricket C2. The closed line surrounding the outline of the thorax 42 is designated as a pattern annotation 50d. The pattern annotation 50d in this embodiment is formed in a rectangular shape. However, the shape of the pattern annotation 50d is not particularly limited and may be any shape.
[0056] According to this embodiment, the pattern annotation 50d makes it possible to distinguish between larvae and adults of the two-spotted cricket C2.
[0057] In addition, among the symbols used in the second and subsequent embodiments, the same symbols as those used in the previous embodiments represent the same components, etc. as those in the previous embodiments, unless otherwise specified.
[0058] The present invention is not limited to the above-described embodiments, and can be applied to various embodiments without departing from the gist of the present invention.
[0059] The breeding device according to the present invention is not limited to those described in Examples 1 and 2, and can be applied to, for example, the breeding device described in JP 2023-037767 A. In the breeding device described in JP 2023-037767 A, when harvesting crickets, the crickets resting on the stop member arranged in the main body frame are dropped from the stop member onto the belt of the belt conveyor, and then the main body frame is removed from the belt conveyor, and the crickets are placed on the belt. In this way, the training image data and test image data according to the present invention can be obtained by photographing the crickets placed on the belt with a camera.
[0060] In addition, when harvesting crickets, the crickets that are resting on the resting members arranged in the main frame are dropped from the resting members onto the belt of the belt conveyor, and then the crickets are stored in a harvesting box at the end of the belt conveyor. By photographing the crickets stored in the harvesting box with a camera, training image data and test image data according to the present invention can be obtained. [Explanation of symbols]
[0061] 1: insect recognition model generation device, 2: insect production management system, 20: machine learning device, 21: camera, 22: input unit, 23: model generation unit, 24: judgment unit, 25: annotation correction unit, 30: memory device, 31: training image data, 32: test image data, 33: trained model, 41: head, 42: thorax, 43: abdomen, 43b: ovipositor, 44a, 44b: exoskeleton, 46: pattern, 50: annotation, 50a: body annotation, 50b: ovipositor annotation, 50c: wing annotation, 50d: pattern annotation, C1: cricket, C2: two-spotted cricket
Claims
1. An insect recognition model generation device for generating an insect recognition model for recognizing insects, comprising: A machine learning device and a storage device, The storage device includes: Training image data for learning in which the insects are linked by annotations; Test image data for evaluation; A trained model generated by performing machine learning using the training image data and the test image data, The machine learning device includes: an image acquisition unit that acquires the training image data and the test image data; an input unit to which the annotation to be assigned to the training image data is input; A model generation unit that generates the trained model by performing machine learning using the training image data; a determination unit that determines whether the insect included in the test image data is recognized when the test image data is determined by the trained model; An insect recognition model generating device comprising: an annotation correction unit that corrects the annotation if the judgment result of the test image data by the judgment unit is incorrect, and maintains the annotation if the judgment result is correct, based on the judgment result of the test image data by the judgment unit.
2. The insect recognition model generating device according to claim 1 , wherein the annotation is performed by surrounding an outline of the insect with a closed line.
3. the figure formed by the closed line is a polygon having a plurality of vertices, The insect recognition model generating device according to claim 2 , wherein the annotation includes information regarding relative positions of the vertices of the polygon and information based on the number of pixels of the polygon.
4. The insect recognition model generating device according to claim 1 , wherein the insect is an insect that undergoes incomplete metamorphosis, in which the larvae and adults have similar shapes and the larvae transform into the adult by repeatedly molting without going through a pupal period.
5. The insect recognition model generating device according to claim 4 , wherein the annotation distinguishes between the insect and a shed skin of the insect.
6. The insect recognition model generating apparatus according to claim 4 , wherein the annotation distinguishes between stages of the insect.
7. The insect recognition model generation device according to claim 4 , wherein the annotation distinguishes whether the insect is an adult or a larva depending on whether or not the insect has wings.
8. The annotation is performed by surrounding the outline of the insect with a closed line; The figure formed by the closed line is a body annotation enclosing the outline of the insect's head, thorax, and abdomen; The insect recognition model generation device according to claim 7 , further comprising: a wing annotation surrounding the outer shape of the wing.
9. The insect recognition model generating device according to claim 8 , wherein the body annotation includes information on relative positions of a plurality of vertices of a polygon formed by a closed line, and information based on the number of pixels of the polygon.
10. The insect recognition model generation device according to claim 4 , wherein the annotation distinguishes between male and female insects based on the presence or absence of an ovipositor.
11. The annotation is performed by surrounding the outline of the insect with a closed line; The figure formed by the closed line is a body annotation enclosing the outline of the insect's head, thorax, and abdomen; The insect recognition model generation apparatus according to claim 10 , further comprising: an ovipositor annotation surrounding an outer shape of the ovipositor protruding from the abdomen.
12. The insect recognition model generating device according to claim 11 , wherein the body annotation includes information on relative positions of a plurality of vertices of a polygon formed by a closed line, and information based on the number of pixels of the polygon.
13. An insect production management system comprising the insect recognition model generation device according to any one of claims 1 to 12.
Citation Information
Patent Citations
Multiplication of cricket for feed
JP1998191834A
