Alignment with DF goals (BGI, Platform growth, community)

RAVEN (Real-time Anomaly and Violence Evaluation Neural network) is an AI-driven surveillance platform that uses computer vision and deep learning to detect and prevent violent incidents before escalation. Processing live video via optimized CNN‑LSTM models, it analyzes human movement, interactions, and weapons to identify aggression. Deployable on edge or cloud GPUs, RAVEN delivers real-time inference and instant alerts, threat ratings, and locations through secure dashboards and mobile notifications. A continuous learning loop retrains the model with incident labels to boost accuracy and reduce false alarms. Privacy measures like face blurring and encryption ensure compliance. Modular design integrates with existing camera networks, access control, and alarm systems, making RAVEN suitable for schools, airports, public spaces, and smart cities. By shifting from reactive to predictive monitoring, RAVEN enhances public safety and mitigates violence.

Problem description

Violence kills and injures millions yearly. In 2019, 475,000 homicide deaths (~1,300/day) and 1.25 M violence‑related deaths (~3,425/day) occurred globally. Non‑fatal violence affects 1 in 3 women. Current CCTV is often used after the violence has been perpetuated and misses early cues. RAVEN (Real‑time Anomaly and Violence Evaluation Neural network) uses AI to analyze live video for aggression patterns and weapons, issuing instant alerts to prevent escalation in high‑risk environments.

Proposed Solutions

RAVEN will use real-time video feeds processed on cloud GPUs with CNN-LSTM models to detect aggressive actions, weapons and anomalous behavior. When risk is detected, the system issues low-latency alerts to security personnel. A feedback loop collects incident labels to continuously retrain and adapt the model for diverse environments, ensuring high accuracy and rapid response.