Skip to content

Conversation

dpascualhe
Copy link
Collaborator

@dpascualhe dpascualhe commented Sep 9, 2025

With this PR we extend our image detection frameworks to support

  • YOLO-styled datasets
  • YOLO models exported to torchscript

We include the corresponding notebook tutorial and also update the GUI to support YOLO. GUI should be reviewed in the future to avoid having too many "ifs" depending on the dataset. It should be transparent to the dataset format if the dataset objects are properly defined.

This PR also:

  • Changes the read_annotation returned tuple. Until now, it was returning bbox, label and category ID, but label and category ID were exactly the same. To avoid duplication I got rid of the label parameter.
  • Ignores classes with no GT instance for mAP computation (following Ultralytic's methodology).

@dpascualhe dpascualhe linked an issue Sep 9, 2025 that may be closed by this pull request
@dpascualhe dpascualhe self-assigned this Sep 9, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Add support for YOLO-styled datasets
1 participant