Advances on photo-identification of Risso's dolphin based on computer-aided innovative approaches

Photo-identification of cetaceans is a widely used technique for associating a name to a sighted (or re-sighted) individual. However, two common issues faced by biologists are the high amount of time required to manually process images and the lack of comprehensive information on some species. In this scenario, computer-aided approaches represent the key for improving knowledge on data deficient species and reduce time required to work on large pictures catalogues. In this paper, advances on the photo-identification of Risso’s dolphins automatically exploiting their typical white scars on the dorsal fin is shown. Multiple experiments were conducted to train and validate the proposed approach, as well as a comparison with the state of the art software DARWIN. Results in terms of high accuracy (90%) and reduced computational time (a few seconds per image) shown that this technique can move biologists towards the automated analysis of large image datasets.