Research applies AI techniques for automatic snake recognition
The application of artificial intelligence techniques for the automatic recognition of snake species was the subject of research carried out recently by Pierre Haslee, a student at France’s National Institute of Applied Sciences. The results were presented during his six-month visit to Fundação Getulio Vargas’ School of Applied Mathematics (FGV EMAp), under the guidance of professors Asla Medeiros e Sá and Alberto Paccanaro, in partnership with Paulo Passos, the curator of snake collections at Rio de Janeiro Federal University’s National Museum.
The goal of this research was to conduct experiments to check whether or not it is possible to infer the species of a snake from an image. The project included tests on images of snakes involved in real incidents. In particular, techniques based on convolutional neural networks were applied to large image databases of Brazilian snakes. According to Haslee, the problem becomes particularly relevant in the context of snakebites, when the medical team responsible for treating the patient must choose an antivenom drug by correctly identifying the species involved.
Throughout his stint at FGV, the student had the help of FGV EMAp professor and research supervisor Asla Medeiros e Sá, a specialist in data visualization, who was able to help him present his results with clear graphs and figures. In addition to her, Professor Alberto Paccanaro was of great help due to his years of experience in machine learning research. In FGV EMAp’s postdoctoral research laboratory, Haslee was able to learn and develop his skills further.
Haslee’s research also benefited from Professor Medeiros e Sá’s network of contacts. The student was introduced to Professor Paulo Passos, the curator of snake collections at Rio de Janeiro Federal University’s National Museum, whose vast knowledge of snakes provided access to important data for the research. According to Haslee, “without him the project wouldn’t have existed.”