Automatic Species Identification in Underwater Environments

Knowledge of the geographic distribution and identification of species is essential for the conservation of biodiversity. With advances in technology and greater accessibility of equipment capable of recording underwater, it was possible to obtain data efficiently. However, it leads to an immense volume of information collected, which requires exhaustive manual processing that requires label, time and money. That is why the creation of tools capable of assisting in the monitoring of these species is so important. In this project, we propose to fill a gap in the literature on the development of species monitoring systems in underwater environments. The proposed approach aims not only to identify the existing species in the training set, but also to be able to identify new species that can be registered in that environment. We intend to explore several methodologies using clustering, dissimilarity and convolutional neural networks. We will also present a new set of public data from high resolution underwater videos that will be available to the scientific community.

Faculty Supervisor:

Alessandro Lameiras Koerich

Student:

Partner:

Federal University of Parana

Discipline:

Computer science

Sector:

Artificial Intelligence; Ocean Tech; Life Sciences (not health)

University:

École de technologie supérieure

Program: