Model trained using 1739 train images and 500 test images.
Converted coco label format to yolo(darknet) format using this frageneral_json2yolo.py in here
Model trained on YOLOv5l pretrained model.
No of epochs for training the model: 4
Model performance on validation dataset:
Class Images Labels Precision Recall
all 501 4003 0.748 0.696
person 501 2654 0.744 0.668
car 501 1349 0.752 0.725
Trained model can be downloaded from this link
install packages required :
pip install -r requirements.txt
To run inference :
python detect.py --source check.jpg --weights last.pt
The output image will be saved as result.jpg
