⚡ EDGE-AI SURVEILLANCE COMMAND CENTER
A lightweight, low-latency, cloud-optimized computer vision ecosystem that processes live client-side browser video streams via WebRTC to detect spatial threats instantly using YOLOv8.
⚙️ Key Architecture
Standard AI web integrations usually suffer from severe lag because video capture and deep learning inference overwhelm a single server thread. This ecosystem completely eliminates hardware bottlenecks by leveraging an asymmetric Client-Server WebRTC Pipeline:
- 🌐 Asynchronous WebRTC Frontend: Captures the user’s camera feed directly inside the web browser using lightweight client-side JavaScript. This offloads all video hardware dependencies entirely from the server.
- 🛡️ Collision-Locked Ingestion: Uses a smart concurrency guard (
isProcessingFrame) to drop overlapping frame transits, ensuring data packets never crowd the network layer.
- 🧠 Optimized ONNX Inference Engine: Executes target tracking via a highly compressed ONNX serialization layout. This keeps memory usage well under 250 MB RAM, rendering bounding boxes instantly even on shared, single-core cloud micro-containers.
🎨 UI Features
The control room dashboard transforms raw matrices into an interactive cyber-defense layout:
- Cyberpunk Command Grid: A high-contrast dark user interface built using custom CSS variables for low eye strain during long-term monitoring.
- Dynamic Filter Slider: An integrated JavaScript range slider that dynamically updates AI prediction confidence boundaries on the fly without a system restart.
- Isolated Log Stream: An auto-refreshing log timeline panel that scrolls real-time threat detection logs on a detached poll rate.
🚀 Installation & Setup
1. Prerequisite Environment
Clone the repository, navigate to the target directory, and install the optimized system requirements:
```bash
git clone https://github.com/SnehaKumariArya/Edge-ai-surveillance.git
cd Edge-ai-surveillance
pip install -r requirements.txt