Hi @stark-t , I started with the implementation of the PyTorch_YOLOv4 repository. While the auhotrs provide a yolov4-tiny.cfg file in PyTorch_YOLOv4/cfg folder, I cannot find their corresponding weights file in the README table with download links.
Maybe I got a bit stuck in my search buble. Could you check as well?
FYI:
meanwhile, I managed to set the environment for PyTorch_YOLOv4 and I currently let a training job run using the yolov4.cfg and thier corresponding yolov4.weights provided in the README. This is for a 640 x 640 image size.
It was a bit buggy to set the environment and it was not smooth as for yolov5. For example, the file requirements.txt missed some of the needed dependencies, like PyYAML & scipy. Then I had to install older versions of numpy and that also needed special modules from the software tree on the cluster. Fingers crossed that now will work.
Hi @stark-t , I started with the implementation of the PyTorch_YOLOv4 repository. While the auhotrs provide a yolov4-tiny.cfg file in PyTorch_YOLOv4/cfg folder, I cannot find their corresponding weights file in the README table with download links.
I looked through the issues, but didn't find much helpful info: https://github.com/WongKinYiu/PyTorch_YOLOv4/search?q=tiny+weights&type=issues
Maybe I got a bit stuck in my search buble. Could you check as well?
FYI: meanwhile, I managed to set the environment for PyTorch_YOLOv4 and I currently let a training job run using the yolov4.cfg and thier corresponding
yolov4.weights
provided in the README. This is for a 640 x 640 image size.It was a bit buggy to set the environment and it was not smooth as for yolov5. For example, the file requirements.txt missed some of the needed dependencies, like
PyYAML
&scipy
. Then I had to install older versions ofnumpy
and that also needed special modules from the software tree on the cluster. Fingers crossed that now will work.