Closed Pickymarker closed 1 year ago
no I should not
v7 will give no boost at the same flops
v5 (current in DFLive):
v7:
also converting to onnx is hard manual job
anyway there is no significant boost to spend the time for this.
YoloV5 is very good for real time apps
is there any improvements that can do done to deepfacefacelive for amd because amd needs more improvement on deepfacelive it can get get but is not getting i would like opencl with hsa compute if not then directml if not also then here are the models someone posted on baidu https://pan.baidu.com/s/1oIj3QgOohKpHGwJDxNvDsw#list/path=%2Fsharelink3574865391-188501854105692%2Fonnx_havepost_models&parentPath=%2Fsharelink3574865391-188501854105692 i would download them but i can't download stuff from baidu because i can't get a account
you should update yolov5 to yolov7 https://github.com/derronqi/yolov7-face and this has a 640 model which deepfacelive i think needs all it needs is to be converted from .pt to onnx