WildEye: Automated Wildlife Monitoring Using Camera Traps, Deep Learning and Real-Time Alert Systems

Abstract

Wildlife conservation demands rapid, accurate, and scalable monitoring of animal species and human activities across vast forest territories. Manual review of camera-trap imagery is labour-intensive, error-prone, and incapable of realtime response. This paper presents WildEye, an AI-powered wildlife monitoring system that integrates MobileNetV2- based deep-learning classification with YOLOv8 real-time object detection inside a Flask web application. MobileNetV2, fine-tuned via transfer learning on a Kaggle wildlife dataset, classifies eleven species (cheetah, elephant, giraffe, hyena, leopard, lion, panda, rhinoceros, tiger, wolf, zebra) with 94.7 % model accuracy and a mean confidence of 95.8 %. YOLOv8 processes live webcam frames and, on detecting a human presence, dispatches an SMTP email alert to the designated wildlife officer. The system provides a responsive web dashboard, species profile library, image gallery, and RESTful API endpoints. Experimental results confirm the viability of deploying lightweight deep learning pipelines for conservation technology.