search

CVHub520 / X-AnyLabeling

Effortless data labeling with AI support from Segment Anything and other awesome models.
GNU General Public License v3.0
3.01k stars 342 forks source link

Feature Request: Option to Preserve Existing Annotations During Auto-Labeling #455

Open peters opened 1 week ago

peters commented 1 week ago

The auto-labeling feature currently overwrites existing annotations, which can disrupt workflows where manual and automatic annotations coexist. To address this, I propose adding a checkbox labeled "Preserve existing annotations" in the auto-labeling settings. When checked, this option should allow the application to skip auto-labeling for data points that already have annotations, preserving any manual input while still labeling unannotated data.

CVHub520 commented 1 week ago

Hi, @peters:

Thank you for your feature request. It's a great suggestion and we'll consider adding it to our roadmap.

The idea of allowing users to preserve existing annotations during auto-labeling is very useful for workflows where manual and automatic annotations are used together. We'll work on incorporating this feature into a future update of X-AnyLabeling.

Please stay tuned for updates, and if you have any other suggestions or issues, feel free to let me know.

Best regards, CVHub

CVHub520 commented 1 week ago

Great news! We’ve just added a ‘Preserve existing annotations’ checkbox to the auto-labeling settings. Now you can easily coexist with manual and automatic annotations without overwriting. Feel free to pull the latest code and give it a try!

peters commented 1 week ago

Wow, that was fast! :dancers: I've pulled the latest code and it works as expected. Great work! Thank you :rocket:

peters commented 1 week ago

When switching between models, existing labels are not preserved. Restarting the application temporarily resolves the problem, but the issue reoccurs when changing models again.

Steps to Reproduce:

Open the application and load a model. Add labels to the data. Switch to a different model. Observe that the previously added labels are not preserved.

Expected Behavior: Labels should be preserved when switching between models.

Actual Behavior: Labels disappear when changing models.

Temporary Workaround: Restarting the application restores the labels, but they are lost again upon changing models.