🚗 Vision-Based ADAS Projects Repository
This repository contains the final project submissions developed by mentees as part of the Vision-Based ADAS (Advanced Driver Assistance Systems) learning track.
Each project implements a camera-based ADAS perception pipeline operating on real driving video data, focusing on lane detection, vehicle detection, spatial reasoning, and safety logic, similar to early-stage ADAS prototypes used in industry.
📌 Project Overview
The goal of the final project was to design and implement a vision-only ADAS system that processes continuous front-facing driving video and produces a unified, annotated output demonstrating:
Road lane understanding
Vehicle awareness using object detection
Region-based spatial reasoning using masking
Rule-based safety and warning logic
Integrated real-time visualization
The emphasis of the project was system design and integration, not model training.
🧠 Technical Scope & Constraints
Object Detection: All projects use a pretrained YOLO model for vehicle detection. Training or fine-tuning of models was explicitly not part of the task.
Lane Detection: Implemented using classical image processing techniques such as edge detection, region masking, and line estimation.
Dataset Usage: Publicly available front-facing driving videos (e.g., dashcam-style datasets) were used strictly as input video data, not for training.
Evaluation: All systems were tested on the same fixed video(s) to ensure consistent comparison across submissions.
🧩 Expected System Capabilities
Each project in this repository implements the following components:
Detection and visualization of road lane boundaries and lane area
Vehicle detection using YOLO with bounding box overlays
A masked Region of Interest (ROI) ahead of the vehicle
Counting of unique vehicles entering the ROI over time
Forward collision risk estimation using rule-based logic
Lane departure warning based on lane center deviation
A single unified output video combining all visualizations
📁 Repository Structure
Each mentee has committed their work in a separate folder, following a consistent structure:
├── mentee_name/ │ ├── src/ # Source code │ ├── models/ # YOLO weights / configs (if applicable) │ ├── input_video/ # Test video(s) │ ├── output_video/ # Annotated ADAS output │ ├── README.md # Project-specific explanation │ └── requirements.txt # Dependencies
Note: Folder names and exact structure may vary slightly based on individual implementation choices.
Each mentee has included a project-level README inside their folder with:
Setup instructions
Dependency installation
Command to run the pipeline
Please refer to the respective folder for execution details.
🎯 Learning Outcomes
Through this project, mentees demonstrated their understanding of:
Real-time video processing
Classical computer vision pipelines
Practical usage of pretrained YOLO models
ROI masking and spatial filtering
Temporal logic and decision-making
System-level thinking for ADAS applications
📎 Disclaimer
These projects are educational prototypes intended for learning purposes only. They do not represent production-ready ADAS systems and should not be used in real driving scenarios.
🙌 Acknowledgements
This repository represents the culmination of a structured learning journey in computer vision for autonomous and driver assistance systems. Great effort by all contributors in bringing together perception, logic, and visualization into complete working systems.