Fixing the current version of the app (+performance issues)
#3
by C-Achard - opened
Model performance issue
One thing that is immediately noticeable is that the code uses an older dlcrnet model, and some DLCLive code that should not be required now that the pytorch branch has a full inference API. One option would be to replace it in a fork and compare how it fares between the two for some images.
Done/fixed
I am working on porting my commits to this repo, however for future reference:
- The build error is due to python 3.13 being used, fixed in #2
- Runtime errors from gradio updates must be fixed mostly in ui_utils.py. See file here
- App.py launch requires a small update, see here
- Tensorflow imports from DLCLive require installing tf, this could be avoided by merging and releasing a version with this dlclive PR : DLCLive#171
- "theme" parameter should move to launch() in app.py
C-Achard changed discussion title from Fixing the current version of the app to Fixing the current version of the app (+performance issues)