search

cvat-ai / cvat

Annotate better with CVAT, the industry-leading data engine for machine learning. Used and trusted by teams at any scale, for data of any scale.
https://cvat.ai
MIT License
12.59k stars 3k forks source link

Feature Request: Show detection confidence on annotation page and allow users to filter annotations by a confidence threshold #5844

Open saifkhichi96 opened 1 year ago

saifkhichi96 commented 1 year ago

I would like to request a new feature that would greatly improve the annotation experience for users: the ability to show detection confidence on the annotation page and filter annotations by a confidence threshold.

Currently, after using automatic annotation, it can be time-consuming to manually check and fine-tune every single annotation. However, if we could quickly browse through annotations where the network wasn't very confident, we could save a lot of time and effort.

I believe that most detector networks already return a confidence score in the serverless functions, and this score is not being used anywhere in the code right now. By displaying this confidence score on the annotation page, users would be able to easily identify annotations that may need further manual review.

In addition, allowing users to filter annotations by a confidence threshold would make the annotation process even more efficient. For example, users could choose to only view annotations with a confidence score above a certain threshold, which would help them focus on the most reliable detections.

Overall, I think this feature would be a valuable addition to CVAT and would greatly improve the annotation workflow for users. Thank you for considering this request.

nmanovic commented 1 year ago

@saifkhichi96 , thank you for the feedback. The feature request is very reasonable. We will add the functionality in the future. If you can help us and contribute, it will be awesome.