WildEye: A Real-Time AI-based Surveillance System for Human-Wildlife Conflict Mitigation

Poaching and Human-Wildlife Conflict (HWC) are two great threats to biodiversity and rural communities due to the fact that both affect the ability of the community and environment to survive. To mitigate these challenges, the WildEye framework introduces an automated, high-precision intelligence system designed for real-time wildlife monitoring. By integrating deep learning-based object detection with low-latency communication protocols, the architecture provides a scalable solution for identifying high-priority species in diverse ecological environments. Using Ultralytics' YOLOv8 deep learning-based detection model, we trained our system on a vast amount of data of at least 27 different animal species which enables the model to recognize animals apart from wild and domestic animals by learning rich feature sets through the training data. Additionally we developed a custom Logic Filter using the Python programming language that restricts any visual alerts or notifications to five species of the highest priority in terms of protected status: Lions, Tigers, Leopards, Elephants, and Deer. After the model has been confirmed for functionality and accuracy, it will query a local animal IUCN conservation type database (e.g., endangered species) and, once this has been retrieved, the model sends this metadata via HTTP POST to a Node.js back-end server for storage in a MongoDB Atlas cloud database. Additionally, we used Socket.io (WebSockets) in place of a traditional polling approach to relay alerts to a real-time dashboard that is based on React.js. The experimental results show that training on a larger dataset reduces the chance that false positives would occur. The architecture of using WebSockets to deliver alerts provides forest officials with contextualized alerts at sub-second latency.