🛡️ ForensVision

ForensVision is an advanced forensic video analysis system that identifies violent behavior and weapons in videos. It integrates a Python backend running machine learning models with a modern Next.js frontend, making it easy to upload videos and view analysis results instantly.

✨ Features

• 🎥 Analyze uploaded videos for violence and weapon detection
• ⚡ Fast and accurate AI-based processing
• 🚔 Designed for forensic investigators, security professionals, and law enforcement

---

🚀 Getting Started

⚙️ Backend Installation

1. Clone and navigate to backend

   git clone https://github.com/ryshel-jasmi/ForensVision.git
   cd ForensVision/backend

2. Create virtual environment

   python -m venv venv
   venv\Scripts\activate     # On Windows
   
   # source venv/bin/activate   # On Linux/macOS

3. Install dependencies

   pip install -r requirements.txt

4. Start the backend server

   python main.py

   Server will start at http://localhost:8000

🖥️ Frontend Installation

1. Navigate to frontend directory

   cd ForensVision/frontend

2. Install dependencies

   npm install

3. Start the development server

   npm run dev

   Frontend will be available at http://localhost:3000

---

⚙️ Environment Configuration

# Backend Configuration
CORS_ORIGINS=http://localhost:3000
MAX_FILE_SIZE=500MB
MODEL_DEVICE=auto

# Frontend Configuration
NEXT_PUBLIC_API_URL=http://localhost:8000

---

🤖 AI Models

Violence Detection Model

• Architecture: MoBiLSTM (Mobile + Bidirectional LSTM)
• Components:
  • LSTM for temporal analysis
  • YOLO for object detection
  • ResNet-50 for visual feature extraction
• Input: Video frames with person tracking
• Output: Violence probability with component scores

Weapon Detection Model

• Architecture: YOLOv8n (Nano variant for speed)
• Training: Custom dataset with multiple weapon types

Model Paths

Models are automatically loaded from the models/ directory:

• Violence: models/violence_detection/MoBiLSTM_violence_detection_model.h5
• Weapons: models/weapon_detection/weapon_detection.pt