vehicle-and-traffic-sign-detection-depended-on-yolov5-and-cascade-classifier

I use both yolov5 and cascade classifier to detect vehicle and traffic sign based on bdd100k dataset. The reference time of yolov5 is about 17ms much better than cascade classifier., which is enough for real-time detection.

cascade classifier

Cascade classifier file to detect cars, pedestrian, stoplights, two wheelers.

data

.yaml file for datasets path

datasets

The bdd100k datain yolo's format. The structure of datasets is same with https://github.com/ultralytics/yolov5.

test_video

Some videos for testing. The file is to big.

weights

yolov5n.engine for inferencing on tensorrt yolov5n_4000_640.pt is trained from 4000 pictures of 640x640 pixels yolov5n_6000_640.pt is trained from 6000 pictures of 640x640 pixels yolov5n_8000_416.pt is trained from 8000 pictures of 416x416 pixels yolov5s_3000_640.pt is trained from 3000 pictures of 640x640 pixels yolov5s_5000_640.pt is trained from 5000 pictures of 640x640 pixels

convert2yolo.py

convert bdd100k labels to yolo

car_detection_yolov5.py

The main program to detect vehicles by yolov5. Change your own parameters in parameters().

car_detection_cascade_classifier.py The main program to detect vehicles by cascade classifier. Change your own parameters in parameters().

Yutong-gannis/vehicle-and-traffic-sign-detection-depended-on-yolov5-and-cascade-classifier