VeSTIS: A Versatile Semi- Automatic Taxon Identification System from Digital Images

Abstract

In this work we present a flexible Open Source software platform for training classifiers capable of identifying the taxonomy of a specimen from digital images. We demonstrate the performance of our system in a pilot study, building a feed-forward artificial neural network to effectively classify five different species of marine annelid worms of the class Polychaeta. We also discuss on the extensibility of the system, and its potential uses either as a research tool or in assisting routine taxon identification procedures.