AI-CoachGuard: The Smart Eyes Inside Every Compartment and Station

Abstract

In India, where unauthorized entry of men into women-only railway compartments often leads to discomfort, harassment, and safety risks, ensuring passenger security remains a major concern. Existing measures such as CCTV surveillance and manual patrolling are largely reactive and struggle to distinguish between unintentional entries and genuine threats, resulting in delayed responses. To overcome this, we propose AI-CoachGuard: The Smart Eyes Inside Every Compartment and Station, an AI-powered framework for real-time surveillance and proactive intervention in women’s compartments. The system integrates facial detection and gender classification using a fine-tuned ResNet18 model to accurately identify unauthorized males, while a Mediapipe-based gesture recognition module detects predefined distress signals, such as SOS gestures, through hand landmark extraction and geometric analysis. A dual-verification mechanism ensures alerts are triggered only when both unauthorized male presence and distress gestures are detected, thereby reducing false positives from staff, vendors, or accidental entries. On detection, AI-CoachGuard sends instant SMS alerts with compartment ID, timestamp, and location to railway authorities and simultaneously updates a live monitoring dashboard for event logging and forensic review. Designed for deployment on real-time CCTV feeds, the framework enables discreet passenger signaling and immediate authority response, with future extensions including IoT-based integration, crowd density monitoring, and stampede prevention for enhanced railway safety.