This paper presents a new dataset for fine-grained visual classification (FGVC) of fish species in their natural environment. It contains 794 images of 12 different fish species collected at the Adriatic sea in Croatia. All images show fishes in real live situations, recorded by high definition cameras. Remote and diver-based videography is used by a growing number of marine researchers to understand spatial and temporal variability of habitats and species. The required large numbers of independent observations necessitate the development of computer vision tools for an automated processing of high volumes of videos featuring high fish richness and density. As baseline experiment, we are using CNN features [1] and a linear SVM classifier and achieve an accuracy of 66.78% on our dataset.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.