How do I port a yolov4 or yolov5 to CoreML

I am using turicreate to train a custom object detection model, however it only supports yolov2. Has anyone tried to port v4 or v5 to CoreML . Is their a utility to do this? I have a couple of v4 PyTorch examples I was going to train then try to do this.
Up vote post of rkey3224@gmail.com Down vote post of rkey3224@gmail.com
2.0k views
Posted by
rkey3224@gmail.com
Reply
Add a Comment

Replies

CoreML is not a GPU computing library is an ML library accelerated on GPU, so it seems similar, but as far as I know, it is not like OpenCL. However, there is a piece of good news.

There is a repository for OpenCL and macOS, and it supports YOLOv2, YOLOv3, YOLOv4, and YOLOv5. It works beautifully and fast, and I did it for myself and my macOS and shared.

This is an open repository you can find at https://github.com/sowson/darknet. If you have time and want to participate, the SWITCH activation function is not implemented yet, so you may play with C and add a PR :-).

Thanks! and Enjoy!

Posted by
sowson
Up vote reply of sowson Down vote reply of sowson

  • There is one more thing, sorry I forgot to mention a live demo at https://igpu.io where you put any image and see results after one button click. Thank you! ;-).

    —
    sowson
Add a Comment