Detection of rare events in videos of wild animals

MARINE, a computer vision model trained on fish predator attacks

Zsófia Katona

MSc Information Studies, Data Science track, University of Amsterdam

Institute of Biodiversity and Ecosystem Dynamics

---

Contents

The thesis concerns action recognition to identify fish predator attacks in videos of marine habitats.

This repository includes the following directories:

Implementation

Includes the implementation of all models and experiments used in the thesis. More information on how to run the project is offered in the directory's README file.

Tutorials

Includes completed notebooks of tutorials related to techniques used in the study.

Programming exercise

Includes the implementation of a programming exercise aimed to kick-off the learning process of the technical aspects of the project.

Figures

Includes figures used in the thesis and on the GitHub page.

---

Acknowledgment

This project was built with the kind help of the supervisors of the thesis, dr. Fatemeh Karimi Nejadasl, dr. Ben Martin, and dr. Sahand Mohammadi Ziabari, and using datasets provided by the Institute for Biodiversity and Ecosystem Dynamics at the University of Amsterdam.