This project implements an advanced object detection and decision-making system for autonomous vehicles. It focuses on enhancing vehicle performance in rural and unfamiliar environments through three main tasks:
- Detecting known objects in video footage
- Detecting novel objects in video (research stage)
- Making collision-avoidance decisions during navigation
Video Demo: [[Video link]]
- Utilizes YOLOv10-Small for efficient, real-time object detection
- Implements specialized crosswalk detection
- Includes traffic light color classification
- Makes real-time collision avoidance decisions based on detected objects
- Python 3.x
- OpenCV
- Ultralytics YOLO
- NumPy
- traffic-light-classifier
- Clone this repository
- Install required packages:
pip install opencv-python ultralytics numpy traffic-light-classifier - Download the required YOLO models:
- YOLOv10-Small model
- Crosswalk detection model
- Prepare your input video
- Run the main script:
python detect.py - Enter the path to your input video when prompted
- Specify the output video name
main.py: Main script for video processing and decision makingmodels/: Directory containing YOLO model weightsstop.txt: List of objects that trigger a stop actionslow_down.txt: List of objects that trigger a slow down action
- YOLOv10 by Ultralytics
- Crosswalk detection model by xN1ckuz
- Traffic light classifier by Shashank Kumbhare
- Implementation of novel object detection using cooperative foundational models and zero-shot detection techniques
Vo Thanh Nghia - vothanhnghia270604@gmail.com
Project Link: [Project link]