BattleEye: A Lightweight CNN Model for Real-Time Military Vehicle Recognition on Low-Power Devices

This paper presents BattleEye, a lightweight convolutional neural network (CNN) architecture specifically designed for real-time military vehicle recognition in airreconnaissance drone footage. BattleEye achieves better performance compared to the YOLOv5 nano model, offering 20% faster inference while maintaining comparable precision, and is optimized for deployment on low-power devices like the Raspberry Pi 3B. The model was trained on a dataset of 5,000 images, enabling it to classify both broad and specific categories of military vehicles. Comprehensive testing on both high-performance PCs and low-power hardware demonstrated BattleEye's ability to deliver efficient and accurate object detection in resource-constrained environments. Additionally, the integration of bounding box clustering further enhances operational efficiency and situational awareness for drone operators.