Detection and tracking of traffic signs, traffic signals and driving lanes using neural networks and computer vision algorithms. Documentation in the form of a Technical paper
This repository is divided into 3 sections:
- Lane detection: Contains work related to driving lane detection only.
- YOLOv4-tiny training: Contains work related to traffic signs and signals detection and recognition using YOLO.
- Tensorflow YOLOv4 with lane detection: Contains final work where the above two things are merged into a single system with Tensorflow framework.
-
First clone this repository and nagivate your directory to the cloned repository.
-
Run-
For Non-GPU:
$ pip install -r requirements.txtFor GPU:
$ pip install -r requirements-gpu.txt-
Navigate to Tensorflow YOLOv4 with lane detection directory
-
Run-
$ python3 detectvideo_lane.pyList of Arguments-
| Argument | Description |
|---|---|
| weights | path to weights file |
| size | resize images to |
| tiny | yolo or yolo-tiny |
| model | yolov3 or yolov4 |
| video | path to input video |
| iou | iou threshold |
| score | score threshold |
| output | path to output video |
| dis_cv2_window | disable cv2 window during the process |
Example commands with arguments-
$ python3 detectvideo_lane.py --weights yolov4-tiny-416 --size 416 --tiny --model yolov4 --video test.mp4 --score 0.50 --output out.avi
$ python3 detectvideo_lane.py --weights yolov4-416 --size 416 --model yolov4 --video test.mp4 --score 0.50 --output out.avi --dis_cv2_window