Re-Identification of Rats with Transfer Learning

Abstract

The study of animal behavior in laboratory experiments is key in ethology, ecotoxicology, neuroscience and other fields. Although modern studies use computer imaging techniques, current solutions cannot preserve the identity of multiple individuals in social experiments. Thanks to the use of Transfer Learning we seek to overcome this limitations while maintaining the effectiveness of Deep Learning and reducing its computational times. With this technique we achieved promising results in the re-identification of rats after an occlusion process