search

rdemets / napari-yolov5

Other
7 stars 1 forks source link

Upgrade to yolov8? #4

Open jo-mueller opened 11 months ago

jo-mueller commented 11 months ago

Hi @rdemets ,

is this still maintained? I was wondering whether it would be of interest to upgrade the plugin to the newer yolov8? Ouf of several improvements, it uses pytorch which I find a bit easier to set up on most machines. Also, the API of the newer ultralytics package is very simple to use.

Best, Johannes

rdemets commented 11 months ago

Hi @jo-mueller ,

It is still kind of maintained when I get error, but I wasnt thinking about upgrading it. Its true that I am not stuck with an "obsolete" method. The group I worked with who requested this plugin are still working using it, so I might have to create a duplicate to be sure their needs is still fulfill, but I can explore the idea.

Thanks for the suggestion.

jo-mueller commented 11 months ago

Hi @rdemets , I didn't mean to imply YoloV5 being obsolete. I only prefer working with pytorch rather than tensorflow as it seems to be a bit easier to install and use, overall. As for YOLO itself, I think it should be backwards compatible to a degree so that any annotations/projects done with YOLO should still be possible to be done with the newer versions - you just don't have to code the whole training loop and all the boilerplate yourself anymore :)

I do agree though that it's probably not worth upgrading from your side as long everything about this plugin is still functional in its current state.