An advanced Advanced Driver Assistance System (ADAS) designed specifically for mobile browsers. This project utilizes a Heterogeneous Architecture (combining Python/Streamlit with JavaScript/WebAssembly) to provide real-time safety analytics directly on the device, avoiding the latency of traditional server-side processing.
This application serves as a prototype for investigating Computer Vision efficiency on Edge Devices. It specifically targets:
- Lane Departure Warning (LDW): Real-time pixel-intensity mapping for white line tracking, visualized with Blue AR Overlays.
- Pedestrian Collision Warning (PCW): COCO-SSD MobileNetV2 inference to identify localized safety zones.
- Road Sign Recognition: Dynamic identification of regulatory signage such as Stop Signs and Traffic Lights.
- Zero-Latency AR: Employs
{desynchronized: true}WebGL rendering to eliminate visual input lag. - Hybrid Execution: Uses Python for the application interface and JavaScript/WASM for high-frequency (30 FPS) vision tasks.
- Resource Optimized: Implements a Priority Queue where lightweight tasks run every frame, while heavy AI inference runs every 7th frame to prevent device overheating.
To ensure the application runs correctly on the Streamlit server, you must create a file named requirements.txt in your main folder and include the following libraries:
- Streamlit (version 1.35.0 or higher)
- Pandas and Numpy for data handling
- Pillow for image processing compatibility
Because the application processes video and graphics, the hosting server requires specific Linux-level headers. Create a file named packages.txt in your root directory and add the necessary graphics libraries (such as freeglut3-dev, libgl1, and libpng-dev) to allow the server to render the vision components correctly.
- Deployment: Upload your source code to a GitHub repository and connect it to Streamlit Cloud.
- Hardware Setup: Mount your smartphone on a vehicle dashboard with a stable, unobstructed view of the area.
- Initialization: Open the provided URL in a mobile browser (Chrome or Safari) and tap the "START SYSTEM" button. You must grant Camera and Audio permissions for the AI to function.
- Visual Indicators:
- Blue Overlays: Indicates that the system has successfully locked onto White Road Lines.
- Red "BRAKE" Alert: Appears instantly if an object or person enters the central collision zone.
- Top-Right Indicator: Displays recognized road signs in real-time.
Copyright (c) 2026 Asari-Rashidi Energy Model Research Group
This software is provided under the MIT License. This allows for free use, modification, and distribution, provided that the original copyright notice and this permission notice are included in all copies or substantial portions of the software. The software is provided "as is," without warranty of any kind.
- Core Model: COCO-SSD (TensorFlow.js)
- Vision Engine: Custom Pixel-Intensity Logic (White-Line Optimized)
- Lead Developer: Asari Rashidi
- Application: Real-world Road Testing