Conversation
Squashed from two prior commits that transiently added and removed ~2.8GB of demo video/data files, which were bloating history without ever being part of the final tree.
demo1/ and similar run directories (videos-raw, videos-labeled, pose-2d, pose-3d, calibration, *.avi) are pipeline outputs, not source-controlled assets -- committing them previously bloated history with ~2.8GB of demo video data.
|
Just on a first pass, there appear to be a lot of changes here that I would think are not needed (e.g. a CUDA backed PyAV loader for the annotator; but there are numerous others). I would expect that in addition to the existing commits on Could you write up a summary of the changes? In addition to the clearly needed modifications to the SLEAP and LP backend files, the rest of the changes should be justified. e.g., "add X method to |
|
A good starting point would be just changes required to add SLEAP or LP without any changes to the TUI. |
|
Yea, i think would be because of the label speed when opening the test demo with the visualizer. There were quite abit of backend changes to take make for sleap guis to luach in the pixi environemnt.bu tyea, I will try to compile a doc on the changes |
|
Are you using the visualizer on a remote machine? None of the backends should launch or even install their respective GUIs in our environment. Labeling is done in a Cheese3D GUI that writes to a backend-agnostic format. The backends will translate back and forth from that format. |
|
Nope, I'm running this locally. I wasn't trying to have Cheese3D launch the SLEAP GUI. I thought it might be useful for people to still access to the native SLEAP, Lightning Pose labeling (very later) workflows (in the pixi environements ), which has iterative prediction/correction labeling. But, I am ok with just have them as minimal back end. Let me know |
|
Let's remove the GUIs from the backends. |
2 participants
see fix_note.txt