Animal Detection in Man-made Environments

摘要: Automatic detection of animals that have strayed into human inhabited areas has important security and road safety applications. This paper attempts to solve this problem using deep learning techniques from a variety computer vision fields including object detection, tracking, segmentation edge detection. Several interesting insights transfer are elicited while adapting models trained on benchmark datasets for real world deployment. Empirical evidence is presented demonstrate the inability detectors generalize training images in their natural habitats deployment scenarios man-made environments. A solution also proposed semi-automated synthetic data generation domain specific training. Code used experiments made available facilitate further work domain.