Add new OCR features, such as re-running OCR on manually annotated local box areas.

[Anleeno]

Search before asking

• [X] I have searched the X-AnyLabeling issues and found no similar feature requests.

Description

Re-running OCR on manually annotated local box areas.

Use case

No response

Additional

No response

Are you willing to submit a PR?

• [ ] Yes I'd like to help by submitting a PR!

[CVHub520]

@Anleeno-Xu, Thank you for your feature request. I'm glad to inform you that the functionality you're looking for is already available in the current version of the software. You can achieve this by making a simple configuration change:

Just take the file anylabeling/services/auto_labeling/ppocr_v4.py as example, locate line 232 and set the replace parameter to False. The line should look like this:

result = AutoLabelingResult(shapes, replace=False)

By setting replace=False, the model will re-run on the manually annotated local box areas without replacing existing annotations. This allows you to perform custom OCR specifically on the areas you've manually defined.

If you have any questions about implementing this change or if you encounter any issues, please don't hesitate to ask. We're here to help ensure you can make the most of this feature.

[Anleeno]

@CVHub520 Thanks. I have tried that, but it doesn't seem to be what I need. What I mean is running OCR on the localized annotation boxes, rather than the entire image, similar to PPOCRLabel. The existing functionality only saves our annotation results when running OCR.

[CVHub520]

Hey @Anleeno-Xu!

Certainly, I understand your request. Currently, this feature is not supported in our product. However, we are considering incorporating it into future development plans to enhance the user experience and meet more specific needs. We appreciate your feedback and will keep you updated on any progress or changes related to this functionality.