omrmhd5/AmanCity

A multimodal AI urban & women's safety platform for Egypt, built to replace fragmented reporting channels. Integrates Tri-Model Computer Vision, live Grok-4 OSINT & predictive ML, consolidating DBSCAN hotspot forecasting, persistent SOS tracking, and Gemini-assisted safe routing into a single proactive ecosystem to protect citizens. # 🛡️ AmanCity – AI-Powered Urban & Women Safety Platform > **Graduation Project (Software Engineering Specialization) — The British University in Egypt (BUE) [June 2026]** Egypt's first fully localized, AI-powered urban and women's safety platform, designed to address the fragmented nature of incident reporting. The platform integrates Computer Vision, Natural Language Processing, and Predictive Machine Learning to detect incidents from crowdsourced media and OSINT data, mapping them in real-time. It significantly improves situational awareness by providing live alerts, predictive risk mapping, and a robust session-based SOS system tailored for personal and women's safety. AmanCity introduces the first active eco-system of safety in Egypt, successfully combining 8 distinct safety capabilities into a single, Arabic-first platform. The system validates crowdsourced data integrity through FFT analysis and transforms raw social and visual data into actionable safety insights. --- ## 🔧 Features ### 📸 Crowdsourced Incident Reporting & AI Gate - Users can report incidents by uploading geo-tagged images, videos, or text. - **FFT-based Synthetic Media Gate**: An inline filter that analyzes high-frequency energy to block AI-generated fake images and deepfakes before inference, ensuring data integrity. ### 👁️ Tri-Model Computer Vision Inference - Three AI models run simultaneously on user uploads using YOLO26 and Random Forest architectures: - **Environmental Classifier**: Detects 7 classes (Fire, Flood, Accident, Road Damage, Damaged Building, Public Issue, Normal). - **Weapons Detector**: Detects 2 classes (Firearm, Cold Weapon). - **Crime Action Recognizer**: Extracts 17-point skeleton poses to detect 11 actions (Assault, Fighting, Robbery, Arson, Shooting, Shoplifting, Stealing, Burglary, Vandalism, Arrest, Explosion). ### 🌐 Social Media Intelligence (OSINT Engine) - Real-time bilingual (Arabic & English) scanning of X (Twitter) using the **xAI Grok-4 API**. - Au …