Re-identification of fish individuals of undulate skate via deep learning within a few-shot context

Individual re-identification is critical to track population changes in order assess status, being particularly relevant species with conservation concerns and difficult access like marine organisms. For this, we propose photo-identification via deep learning as a non-invasive technique discriminate between individuals of the undulate skate (Raja undulata). Nevertheless, accruing enough training samples might be achieve case underwater fish images. We develop novel methodology based on siamese neural network that incorporates statistical fundamentals motivation overcome few-shot context. Our work provides hands-on experience highlights pitfalls when trying apply limited scenario, concerning both data quantity quality, yet providing remarkable results over test set including recaptures, where model capable correctly identifying 70% individuals. The findings this study can strong impact for research teams becoming familiar approaches, it easily extended re-identify other interest from or exploitation point view.