Feature Request: Crowdsourced Error Correction Mechanism on the website

[Roman-]

The COCO dataset occasionally contains annotation errors. A systematic way to report and correct these errors would enhance the dataset's accuracy and utility.

Feature Proposal:
Implement a crowdsourced error correction mechanism, allowing users to report and rectify annotation errors directly.

Key Points:

• User Reporting Interface: Simple UI for error reporting within the dataset platform.
• Error Verification: Process for validating reported errors, possibly automated or manual by maintainers.
• Community Engagement: Encourage user participation in dataset improvement.
• Enhanced Accuracy: Continuous user feedback can improve dataset reliability over time.

Example Errors:
According to my estimates, as of December 2023, around 4% of the images in the COCO dataset contain annotations with at least one misspelled word.

[JasseurHadded1]

There are also many annotation errors in coco-pose dataset! This feature can help reducing these problems!