Article’s

The Military Object Detection System represents a significant advancement in the application of artificial intelligence for defense and surveillance applications. This research presents the development and deployment of a robust, automated solution utilizing YOLOv7 (You Only Look Once, version 7) architecture for detecting military bases and related objects across diverse media formats including static images, recorded videos, and real-time webcam streams. The system is designed to operate securely in offline environments, ensuring strict data privacy and compliance with defense protocols. The project encompasses data preparation with 13 military object classes (Aircraft, Camouflage, Drone, Fire, Grenade, Hand Gun, Knife, Military- Vehicle, Missile, Pistol, Rifle, Smoke, and Soldier), comprehensive model training utilizing state-of-the-art deep learning techniques, and deployment across multiple input modalities. Performance evaluation demonstrates exceptional results with mean Average Precision (mAP) at 0.5 intersection over union reaching 86.9% and precision at 90.1%, indicating strong generalization and reliability. The system delivers real-time detection at over 30 frames per second, making it highly suitable for operational surveillance and threat assessment. This paper presents the complete methodology, architectural details, experimental results, and operational deployment strategies for a practical deep learning-based military object detection system Keywords: YOLOv7, object detection, military applications, deep learning, real-time detection, surveillance, convolutional neural networks, automation