search

talmolab / sleap

A deep learning framework for multi-animal pose tracking.
https://sleap.ai
Other
435 stars 96 forks source link

Loss weight ranges too limited for multi-head models #1971

Open talmo opened 1 month ago

talmo commented 1 month ago

In the GUI, the numeric textbox for adjusting the relative loss weights for the different heads is really not great at the moment.

Issues:

  1. It doesn't let you go below 0.01, but usually we want to move in orders of magnitude (e.g., to reweight supervised classification vs confmaps loss). You can get around this by increasing the loss weight of the other heads to get the equivalent weighting, but it's not ideal.
  2. Clicking on the up/down arrows should x10/x0.1 the value.
  3. You can go negative, which doesn't make any sense for loss scaling.
github-actions[bot] commented 1 month ago

Thank you for opening a new issue!

We appreciate your input and will review it soon.

[!WARNING]
A friendly reminder that this is a public forum. Please be cautious when clicking links, downloading files, or running scripts posted by others.

  • Always verify the credibility of links and code.
  • Avoid running scripts or installing files from untrusted sources.
  • If you're unsure, ask for clarification before proceeding.

Stay safe and happy SLEAPing!

Best regards, The Team