Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Computer vision methods for fish monitoring and inspection
Aquaculture is an important global contributor to the production of seafood for human consumption. Currently the industry is phasing several challenges which demand adaptation of novel technologies and methods to move the production from manual and experience-based to more objective approaches [1]. There is need for objective monitoring and inspection of fish conditions to contribute to better fish health and secure fish welfare.
Keywords: Computer vision, Object detection and Tracking
Current common industry standards are relying on manual assessment of fish status from personnel in fish farms utilizing vision data from feeding cameras and/or manual inspection of sample of the population (50-100 randomly picked fish out of 200.000 from each cage). There is great potential to objectively monitor and inspect the fish population and identify potential change on their behaviour and/or abnormalities on fish (e.g., wounds, missing fins, etc) which can provide more insight on fish conditions and wellbeing [1][2]. This thesis aims to investigate, implement and test methods for fish behaviour and abnormalities identification utilizing visual data obtained from industrial scale fish farms [2]. In addition to access to relevant datasets from stereo camera, there will be support from SINTEF ACE full scale aquaculture laboratory and biologists to obtain further information and inputs on the relevance of the work [3].
References:
[1] Martin Føre et al. “Precision fish farming: A new framework to improve production in aquaculture”. Biosystems Engineering, 2017
[2] Aya Saad, E. Kelasidi, S. Jakobsen, M. Mulelid and M. Bondø, “StereoYolo+DeepSORT: A Framework to Track Fish from Underwater Stereo Camera in Situ”, 17th International Conference on Machine Vision (ICMV), Armenia, 2023
[3] SINTEF ACE. https://www.sintef.no/en/all-laboratories/ace/
Current common industry standards are relying on manual assessment of fish status from personnel in fish farms utilizing vision data from feeding cameras and/or manual inspection of sample of the population (50-100 randomly picked fish out of 200.000 from each cage). There is great potential to objectively monitor and inspect the fish population and identify potential change on their behaviour and/or abnormalities on fish (e.g., wounds, missing fins, etc) which can provide more insight on fish conditions and wellbeing [1][2]. This thesis aims to investigate, implement and test methods for fish behaviour and abnormalities identification utilizing visual data obtained from industrial scale fish farms [2]. In addition to access to relevant datasets from stereo camera, there will be support from SINTEF ACE full scale aquaculture laboratory and biologists to obtain further information and inputs on the relevance of the work [3].
References: [1] Martin Føre et al. “Precision fish farming: A new framework to improve production in aquaculture”. Biosystems Engineering, 2017 [2] Aya Saad, E. Kelasidi, S. Jakobsen, M. Mulelid and M. Bondø, “StereoYolo+DeepSORT: A Framework to Track Fish from Underwater Stereo Camera in Situ”, 17th International Conference on Machine Vision (ICMV), Armenia, 2023 [3] SINTEF ACE. https://www.sintef.no/en/all-laboratories/ace/
- Perform a literature review on current status and challenges in fish farming, object detection, classification and tracking, and image pre-processing techniques used in underwater environments
- Adapt pipeline for object detection, classification and tracking specifically tailored for underwater industrial fish farming scenarios, with the aim to increase robustness and real-time performance
- Assess the performance of the methods and implement the most suitable ones for the given task
- Identify and report on abnormalities on fish and fish behaviour change observed in recorded data
- Perform a literature review on current status and challenges in fish farming, object detection, classification and tracking, and image pre-processing techniques used in underwater environments - Adapt pipeline for object detection, classification and tracking specifically tailored for underwater industrial fish farming scenarios, with the aim to increase robustness and real-time performance - Assess the performance of the methods and implement the most suitable ones for the given task - Identify and report on abnormalities on fish and fish behaviour change observed in recorded data
- Highly motivated student
- Experience with computer vision
- Experience with C/C++ and/or Python programming is desired
- Highly motivated student - Experience with computer vision - Experience with C/C++ and/or Python programming is desired