NNPool in SPIR pipeline for Risso's dolphins identification

Photo-identification is designed to recognize single individuals of a species by exploiting unique and distinctive marks identifiable in one or multiple photographs. The Risso's dolphin's distinctive marks and scars on the dorsal fin proved to be very useful to photoidentify individuals. The aim of this paper is to improve the performances of the photo-identification studies on this species through machine learning techniques. This paper proposes the introduction of NNpool, based on Convolutional Neural Networks, in the SPIR pipeline to identify Risso's dolphins in a new photograph.