Real-time Traffic Signs and Lane Detection

Overview

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.

Installation

• First clone this repository and nagivate your directory to the cloned repository.

• Run-

For Non-GPU:

$ pip install -r requirements.txt

For GPU:

$ pip install -r requirements-gpu.txt

Setup and run

• Navigate to Tensorflow YOLOv4 with lane detection directory

• Run-

$ python3 detectvideo_lane.py

List 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 

References

1. Yolo v4, v3 and v2 for Windows and Linux
2. tensorflow-yolov4-tflite