Incident Lens AI 🔍⚖️

Professional Forensic Video Analysis & Accident Reconstruction Platform

https://youtu.be/QUVeahUrCTg?si=0KiQewMFjjllYqv4

Incident Lens AI is a production-grade application designed for insurance carriers, legal defense teams, and fleet safety managers. It leverages the multimodal capabilities of Google Gemini 3 Pro to transform unstructured video evidence (dashcam, CCTV, bodycam) into legally admissible forensic reconstructions.

Unlike standard video players, Incident Lens AI "reasons" about the footage in real-time, calculating vehicle speeds, inferring traffic signal states from indirect visual cues, and citing specific legal statutes for fault determination.

🚀 Key Features

🧠 Autonomous Reconstruction

• Physics Engine: Automatically calculates vehicle speed ($v=d/t$) using photogrammetry and motion blur mechanics.
• Signal Inference: Deduce the state of occluded traffic lights by analyzing cross-traffic flow and pedestrian behavior.
• Debris Field Analysis: Reverse-engineer impact vectors based on glass shard trajectories and fluid spray patterns.

⚖️ Legal Admissibility

• Search Grounding: Uses the Gemini googleSearch tool to cross-reference observed driving behaviors with real-time case law and vehicle code statutes.
• Chain of Custody: Performs automated authenticity audits to detect video tampering, frame skipping, or deepfake manipulation.
• Reasoning Trace: Displays a transparent, step-by-step logic log (Thinking Mode) before generating final conclusions to ensure explainability.

🔊 Multimodal Synthesis

• Audio-Visual Fusion: Correlates acoustic signatures (tire squeals, horns, impact thuds) with visual frames to determine reaction times precisely.
• Deep Scan ROI: Users can crop specific regions of interest (license plates, signs) for high-resolution targeted forensic queries.

📊 Professional Reporting

• Interactive Dashboard: comprehensive visualization including timelines, fault allocation charts (Pie), and driver risk profiles (Radar).
• PDF Dossiers: Auto-generates specialized reports for different stakeholders:
  • Executive Summary (Claims Adjusters)
  • Technical Reconstruction (Engineers)
  • Legal Brief (Litigators)

---

💎 Gemini 3 Integration

This project showcases the cutting-edge capabilities of the Gemini 3 Pro model via the @google/genai SDK:

1. Native Multimodality: Ingests interleaved high-resolution video frames and raw audio waveforms (WAV) in a single context window to understand the complete scene.
2. Streaming with Tool Use: implementing generateContentStream combined with googleSearch to provide real-time updates while simultaneously querying external legal databases.
3. Structured JSON Generation: The model is prompted to output strictly typed JSON data structures that directly drive the React UI components (Charts, Timelines, Risk Meters).
4. Spatial Understanding: Leverages the model's ability to estimate spatial dimensions (wheelbase length, lane width) for physics calibration without LiDAR data.

---

🛠️ Tech Stack

• Frontend: React 19, TypeScript, Vite
• Styling: Tailwind CSS (Dark Mode Professional UI)
• AI: Google GenAI SDK (@google/genai)
• Visualization: Recharts (Radar/Pie Charts), Canvas API (Annotations)
• Media Processing: Native AudioContext API for waveform extraction
• Reporting: jsPDF for client-side document generation
• Icons: Lucide React

---

⚡ Getting Started

Prerequisites

• Node.js v18+
• A Google Cloud Project with the Gemini API enabled.
• An API Key with access to gemini-3-pro-preview.

Installation

1. Clone the repository

   git clone https://github.com/yourusername/incident-lens-ai.git
   cd incident-lens-ai

2. Install dependencies

   npm install

3. Configure API Key

   • Create a .env file in the root directory (or set it in your deployment environment).
   • Security Note: This is a client-side demo. In production, proxy API calls through a backend.

   API_KEY=your_gemini_api_key_here

4. Run Development Server

   npm run dev

5. Build for Production

   npm run build

---

📖 Usage Guide

1. Upload Evidence: Drag & drop video files (MP4, MOV) or select a "Training Scenario" from the sidebar.
2. Initialize Analysis: Click the "Initialize Analysis" button. The system will extract frames and audio.
3. Monitor Reasoning: Watch the "Live Analysis Stream" to see Gemini's real-time forensic deduction process.
4. Explore Dashboard:
   • Timeline: Scrub through critical events.
   • Liability: Review fault percentages and cited laws.
   • Physics: Adjust friction coefficients in the "Variable Sandbox" to test speed calculations.
   • Chat: Ask "What If" counterfactual questions (e.g., "What if the driver was going 35mph?").
5. Export: Click "Export Case" to download a professional PDF dossier.

---

Author name: Sherin Joseph Roy

email: sherin.joseph2217@gmail.com

website: sherinjosephroy.link

🛡️ License

This project is licensed under the MIT License - see the LICENSE file for details.

Disclaimer: Incident Lens AI is a decision-support tool. All forensic conclusions should be verified by a certified human accident reconstructionist before use in legal proceedings.