Deep-track: A real-time animal detection and monitoring system for mitigating human-wildlife conflict in fringe areas

Abstract

Human-wildlife conflict is a big challenge in the fringe areas where wildlife and humans meet each other often resulting in their harm. For instance, in India, conflicts like Elephants obstructing crops, deer hitting and jumping cars, and Leopard attacking herds have become more and more common in rural areas. This research proposes a novel solution to reduce wild animals’ negative impact on humans by suggesting a field surveillance system that uses camera technology and deep learning for real-time wildlife detection as the main aspects of reducing conflicts. The study was performed in areas dealing with human-wildlife co-existence using video streams captured along the boundaries of the human settlements. The digital video footage is being analyzed for insights per frame, using Visual Geometry Group, VGG16 model, a deep-learning model that has undergone transfer learning, fine-tuned using the Serengeti dataset available on Kaggle to classify and identify the animal species. The animals are also detected in the images and are enclosed in bounding boxes by the VGG16 model, which is implemented with TensorFlow and Keras. The deep-SORT (Simple Online and Real-Time Tracking) algorithm is activated to track and follow the animals in real-time, which makes the process of tracking multiple animals and their movement straightforward. This information is then sent to the local authority, people and commuters via an email or emergency messaging systems (Twilio, MailTrap), alerting them to be vigilant to avoid animal intrusions, protect from animal attacks, crop protection and wildlife conservation effort. During the extensive testing period of the proposed system, the results revealed that the system was able to achieve an accuracy rate of 92.19% with a remarkable precision, recall, and F1-score, thus, proving that the system is accurate in both the detection and classification of wild animals. In conclusion, the proposed system will definitely reduce the human-wildlife conflict and supports human as well as agricultural protection and biodiversity conservation.