Towards Automatic Ethiopian Endemic Animals Detection on Android Using Deep Learning

Nowadays, deep learning became the dominant object recognition technique. Due to its outstanding feature extraction and adaptable capabilities, however, how to ensure speed and accuracy is still a significant task in the area of object detection. In this paper, we provide large-scale animals dataset for the object detection task. The animal's dataset consists six meaningful classes: chilada baboon, red fox, walia ibex, bale mountain vervet, black maned lion, and Somali wild ass. In addition, we utilize a You only look once (YOLO) v3 algorithm for animal detection. YOLOv3 is one of the most extensively used deep learning object detection algorithms; And we have developed user friendly graphical user interface (GUI) based mobile android application. Experiments are performed on our animal dataset that are collected from different positions. The results demonstrate that the proposed approach is highly efficient and effective which provides an accuracy of 97.24% on our animal dataset.