The dataset contains images of the discarded fish on the conveyor belt and annotations. Annotations are prepared in YOLO format, i.e. separate text files, containing fish species label, object bounding box annotation, weight and occlusion level. Annotation per individual fish is written in a separate row of the file.
Additionally, we provide weight file (.pt) for the best performing Detection-Weight2 model.
Research output: Contribution to journal › Article › Academic › peer-review
Sokolova, M. (Creator), Cordova Neira, M. (Creator), Nap, H. (Creator), van Helmond, E. (Creator), Mans, M. (Creator), Vroegop, A. (Creator), Mencarelli, A. (Creator), Kootstra, G. (Creator) (3 Aug 2023). Data underlying the publication: An integrated end-to-end deep neural network for automated detection of discarded fish species and their weight estimation. Wageningen University & Research. 10.4121/a6d5a40e-0358-47cf-9ec1-335df0e4a3c3