This project is an innovative AI-driven surveillance system designed to enhance women’s safety through real-time CCTV monitoring. It detects potential threats, such as a woman being alone at night or surrounded by multiple men, and triggers instant alerts for a rapid response.
Women’s safety in public spaces is a growing concern. we address this issue by leveraging AI-powered analysis of CCTV feeds, allowing authorities to take preventive action before incidents occur.
- 🎥 24/7 Video Monitoring – Continuous surveillance for threat detection
- 👩🦰 Gender Classification – Uses RCNN to identify individuals' gender
⚠️ Real-Time Threat Detection – Analyzes male-to-female ratio, especially at night- 📊 Risk Score Calculation – Factors in time, location, and surroundings
- 🔥 Hotspot Identification – Identifies high-risk zones using historical data
- ✋ Gesture Recognition – Detects distress signals like frantic waving or SOS gestures
- 🚨 Instant Alerts – Sends real-time notifications to authorities
- Python 🐍 – Core development language
- PyTorch 🔥 – Model training & real-time updates
- Mobile VNet 🖼️ – Feature extraction from CCTV frames
- Fine-Tuned YOLO 👤 – Gender classification
- Vision Transformers (ViTs) ⚡ – High-level feature extraction
- MediaPipe 🎥 – Gesture recognition
- CUDA-Enabled GPUs ⚙️ – Accelerated video processing
- Frame Preprocessing – Resizes and normalizes CCTV frames (640×640×3)
- Feature Extraction – CSP-Darknet 53 generates detailed feature maps
- Gender Detection – RCNN classifies individuals with bounding boxes
- Risk Assessment – Computes risk scores based on time, location, and crowd composition
- Hotspot Detection – Identifies high-risk areas
- Alerting System – Sends real-time alerts for detected threats
