This is a AI projects that aims to warn drivers about potential collisions using dashcam video analysis.
To build a real-time collision alert system that analyzes dashcam footage and gives warnings like:
⚠️ "Car CLose"⚠️ "Truck Close"⚠️ "Bus Close"⚠️ "Warning: Pedestrain/Bicycle"
These alerts help drivers stay cautious in complex traffic conditions.
- Dashcam mounted on car windshieled
- Works even when car bonnet is partially visible
- Tested in rainy weather conditions to improve robustness
- Python
- OpenCV
- YOLOv5(pretrained model) for object detection
- Real-time frame wise video analysis
- Detect vehicles and humans from dashcam video
- Warns based on object type and proximity
- Displays alerts dynamically on the video frame
- Capable of running in moderate weather interference (e.g rain)
- Supports videos with bonnet partially in frame
- Improving false-positive filtering using confidence level threshold and bounding box logic
- Adding distance estimation for more accurate warnings
- Planning to integrate audio alerts
- UI/UX polish and documentation underway
this project is being improved regularly by experiementing with differentdetection strategiesand edge cases.
Snehal Dubey (B.Tech CSE student, RTU KOta), Learning AI/ML through hands-on projects like this
If you are into Computer Vision, YOLO, Or AI prjoects like this, feel free to raise issues or suggest improvements. Let's make safe driving smarter together!