Identification of Wild Species in Texas from Camera-trap Images using Deep Neural Network for Conservation Monitoring

Abstract

Protection of endangered species requires continuous monitoring and updated information about the existence, location, and behavioral alterations in their habitat. Remotely activated camera or “camera traps” is a reliable and effective method of photo documentation of local population size, locomotion, and predator-prey relationships of wild species. However, manual data processing from a large volume of images and captured videos is extremely laborious, time-consuming, and expensive. The recent advancement of deep learning methods has shown great outcomes for object and species identification in images. This paper proposes an automated wildlife monitoring system by image classification using computer vision algorithms and machine learning techniques. The goal is to train and validate a Convolutional Neural Network (CNN) that will be able to detect Snakes, Lizards and Toads/Frogs from camera trap images. The initial experiment implies building a flexible CNN architecture with labeled images accumulated from standard benchmark datasets of different citizen science projects. After accessing satisfactory accuracy, new camera-trap imagery data (collected from Bastrop County, Texas) will be implemented to the model to detect species. The performance will be evaluated based on the accuracy of prediction within their classification. The suggested hardware and software framework will offer an efficient monitoring system, speed up wildlife investigation analysis, and formulate resource management decisions.