List of explanations

Fishery AI

Examples of research themes

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Examples of research themes

Research Achievements (Excerpts)

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fisheries × AI

List of explanations

This research aims to utilize AI/IoT technology in the fishing industry to improve operations and advance resource management. We are developing a system that uses fixed cameras to photograph fish caught at fishing ports and automatically recognize the type, size, and number of fish. This system will enable large-scale electronic recording of catch information, leading to more efficient resource surveys. Furthermore, if data can be accumulated continuously, it is expected that it can be used as basic data for future fisheries research. This research is partially supported by JST's ACT-X.

Examples of research themes

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

List of explanations

Fishery AI

Examples of research themes

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Examples of research themes

Research Achievements (Excerpts)

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fisheries × AI

List of explanations

This research aims to utilize AI/IoT technology in the fishing industry to improve operations and advance resource management. We are developing a system that uses fixed cameras to photograph fish caught at fishing ports and automatically recognize the type, size, and number of fish. This system will enable large-scale electronic recording of catch information, leading to more efficient resource surveys. Furthermore, if data can be accumulated continuously, it is expected that it can be used as basic data for future fisheries research. This research is partially supported by JST's ACT-X.

Examples of research themes

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

List of explanations

Fishery AI

Examples of research themes

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Examples of research themes

Research Achievements (Excerpts)

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fisheries × AI

List of explanations

This research aims to utilize AI/IoT technology in the fishing industry to improve operations and advance resource management. We are developing a system that uses fixed cameras to photograph fish caught at fishing ports and automatically recognize the type, size, and number of fish. This system will enable large-scale electronic recording of catch information, leading to more efficient resource surveys. Furthermore, if data can be accumulated continuously, it is expected that it can be used as basic data for future fisheries research. This research is partially supported by JST's ACT-X.

Examples of research themes

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

List of explanations

Fishery AI

Examples of research themes

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Examples of research themes

Research Achievements (Excerpts)

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fisheries × AI

List of explanations

This research aims to utilize AI/IoT technology in the fishing industry to improve operations and advance resource management. We are developing a system that uses fixed cameras to photograph fish caught at fishing ports and automatically recognize the type, size, and number of fish. This system will enable large-scale electronic recording of catch information, leading to more efficient resource surveys. Furthermore, if data can be accumulated continuously, it is expected that it can be used as basic data for future fisheries research. This research is partially supported by JST's ACT-X.

Examples of research themes

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

List of explanations

Fish image synthesis using 3D-CG software

To realize image recognition of catches, a large labeled image dataset is necessary. However, labeling work, especially for instance segmentation, is enormous, and repeatedly outlining numerous relevant regions in images at various fishing grounds is not practical. Therefore, this research develops a method to automatically generate training datasets through image synthesis using a very small amount of labeled data and the 3D-CG software Blender. Compared to the Copy-Paste method, it enables pose transformation, shadow generation, and the expression of three-dimensionality, allowing for the generation of more natural synthetic images, which is a key feature for improving the quality of the training dataset.

Fish body length estimation using Mask Keypoint R-CNN

For the sustainable use of fishery resources, it is crucial to conduct resource surveys using information such as fishing operations and landings, perform resource assessments based on scientific knowledge and objective indicators, and implement resource management according to indicators based on the assessment results. However, basic information necessary for resource assessment, such as the number of fish caught, species, and body length, is often manually measured at each fishing ground. This research develops a system to automatically collect basic information on catches using image recognition with Mask Keypoint R-CNN. Key features include the novel definition of keypoints for catch recognition and the training of the model using only a small amount of labeled data through Copy-Paste Augmentation.

Fishery AI

Examples of research themes

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Examples of research themes

Research Achievements (Excerpts)

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fisheries × AI

List of explanations

This research aims to utilize AI/IoT technology in the fishing industry to improve operations and advance resource management. We are developing a system that uses fixed cameras to photograph fish caught at fishing ports and automatically recognize the type, size, and number of fish. This system will enable large-scale electronic recording of catch information, leading to more efficient resource surveys. Furthermore, if data can be accumulated continuously, it is expected that it can be used as basic data for future fisheries research. This research is partially supported by JST's ACT-X.

Examples of research themes

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Recognition of catches using Mask Keypoint RCNN, automatic generation of catch image data using 3D-CG software, etc.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

Research Achievements (Excerpts)

Fish length recognition using Mask Keypoint R-CNN for fisheries resource management, IPSJ 32nd CDS Workshop, 2021.
Data augmentation method for fish instance segmentation using automatic object generation, Japanese Society for Artificial Intelligence National Conference, 2021.
(Research funding) Development of a general-purpose catch recognition platform for big data in the fisheries industry, JST ACT-X, 2020.12 - 2023.3.

List of explanations

Fish image synthesis using 3D-CG software

To realize image recognition of catches, a large labeled image dataset is necessary. However, labeling work, especially for instance segmentation, is enormous, and repeatedly outlining numerous relevant regions in images at various fishing grounds is not practical. Therefore, this research develops a method to automatically generate training datasets through image synthesis using a very small amount of labeled data and the 3D-CG software Blender. Compared to the Copy-Paste method, it enables pose transformation, shadow generation, and the expression of three-dimensionality, allowing for the generation of more natural synthetic images, which is a key feature for improving the quality of the training dataset.

Fish body length estimation using Mask Keypoint R-CNN

For the sustainable use of fishery resources, it is crucial to conduct resource surveys using information such as fishing operations and landings, perform resource assessments based on scientific knowledge and objective indicators, and implement resource management according to indicators based on the assessment results. However, basic information necessary for resource assessment, such as the number of fish caught, species, and body length, is often manually measured at each fishing ground. This research develops a system to automatically collect basic information on catches using image recognition with Mask Keypoint R-CNN. Key features include the novel definition of keypoints for catch recognition and the training of the model using only a small amount of labeled data through Copy-Paste Augmentation.

Fish image synthesis using 3D-CG software

To realize image recognition of catches, a large labeled image dataset is necessary. However, labeling work, especially for instance segmentation, is enormous, and repeatedly outlining numerous relevant regions in images at various fishing grounds is not practical. Therefore, this research develops a method to automatically generate training datasets through image synthesis using a very small amount of labeled data and the 3D-CG software Blender. Compared to the Copy-Paste method, it enables pose transformation, shadow generation, and the expression of three-dimensionality, allowing for the generation of more natural synthetic images, which is a key feature for improving the quality of the training dataset.

Fish body length estimation using Mask Keypoint R-CNN

For the sustainable use of fishery resources, it is crucial to conduct resource surveys using information such as fishing operations and landings, perform resource assessments based on scientific knowledge and objective indicators, and implement resource management according to indicators based on the assessment results. However, basic information necessary for resource assessment, such as the number of fish caught, species, and body length, is often manually measured at each fishing ground. This research develops a system to automatically collect basic information on catches using image recognition with Mask Keypoint R-CNN. Key features include the novel definition of keypoints for catch recognition and the training of the model using only a small amount of labeled data through Copy-Paste Augmentation.