Animal Detection and Recognition in Day Light Videos Using Deep Learning
Sudharshan Duth P, Dhanyashree AN, P. Chandana · 2024
Understanding animal movements, their ecological behaviors, and interactions during various time intervals of a day is essential for detecting and recognizing animals. This research focuses on animal detection and recognition in different time intervals of a day to address the challenges posed by varied backgrounds, shadows, illuminations, and blurriness in animal videos. In this study, a set of 10 animals were chosen as classes, and experiments were conducted using the YOLOv5 model to enhance the efficiency and accuracy of animal detection during the day compared to other models. Additionally, various deep learning models such as VGG16, VGG19, Faster R-CNN, and YOLOv5 were employed for experimentation to assess their realtime processing capabilities and accuracy. The results showed that the YOLOv5 model outperformed other models, achieving an 85 accuracy rate for detecting and recognizing animals.