Brief description
This dataset has been developed as part of the Australian Research Data Commons Data Discoveries program (https://ardc.edu.au/project/machine-learning-dataset-creation-for-australian-fish-species-from-baited-remote-underwater-videos-bruv/), with the aim to futher advance research into machine learning for the automated detection of fish from video. The dataset was generated from over 3000 videos which were historically analysed with the Event Measure software package and sourced from the Australian Institute of Marine Science (AIMS), University of Western Australia (UWA) and Curtin University of Technology.The dataset is comprised of the following:
- ~80k labelled crops of fish extracted from the videos, from over 500 species, 200 genera and 70 families
- ~45k bounding box annotations (suitable for YOLO,RetinaNet) of fish/no fish across 1800 frames
Lineage
Maintenance and Update Frequency: asNeededCrops of fish were identified using the length measurement information from the Event Measure output files, then cropped and associated with the labels given at time of analysis.
Bounding boxes of fish were generated by random sampling frames from the historical video archive that had associated measurements, and having them analysed on the Amazon Sagemaker Ground Truth platform (https://aws.amazon.com/sagemaker/groundtruth/). The bounding boxes are combined results from multiple observers.
Notes
CreditAustralian Research Data Commons, Data Discoveries program, Machine learning dataset creation for Australian fish species from Baited Remote Underwater Videos (BRUV), Australia.
Bond, T - University of Western Australia (UWA)
Langlios, T - University of Western Australia (UWA)
Harvey, E - Curtin University of Technology
Modified: 12 03 2024
text: westlimit=112.93945312500001; southlimit=-24.126701958681682; eastlimit=137.54882812500003; northlimit=-10.228437266155943
Marrable, D., Tippaya, S., Barker, K., Harvey, E., Beirwagen, S., Wyatt, M., Bainbridge, S., Stowar, M. (2023) Generalised deep learning model for semi-automated length measurement of fish in stereo-BRUVS. Frontiers in Marine Science. 10. https://doi.org/10.3389/fmars.2023.1171625
doi :
https://doi.org/10.3389/fmars.2023.1171625
Marrable, D., Barker, K., Tippaya, S., Wyatt, M., Bainbridge, S., Stowar, M., & Larke, J. (2022). Accelerating Species Recognition and Labelling of Fish From Underwater Video With Machine-Assisted Deep Learning. Frontiers in Marine Science, 9. https://www.frontiersin.org/articles/10.3389/fmars.2022.944582
uri :
https://www.frontiersin.org/articles/10.3389/fmars.2022.944582
Codebase used to generate fish crops from event measure files
uri :
https://github.com/marrabld/open_fish_classifier
GitHub Repository - OzFish - Public dataset of Australian fish species for advancing machine learning research
- global : 38c829d4-6b6d-44a1-9476-f9b0955ce0b8