Two versions of annotations within task?

[chekhovana]

I plan to organize my data annotation pipeline in the following way:

• use model predictions to pre-annotate my data
• manually correct these annotations

Is there a way to keep 2 versions of annotations within single task? By comparing model predictions and manually corrected annotations it would be possible to:

• validate model accuracy
• calculate the price of annotator work

I tried to adapt Honeypot feature, but it doesn't suit this usecase, as it's impossible to create second job with 100% images.

[bsekachev]

Is there a way to keep 2 versions of annotations within single task?

No, there is not.

@zhiltsov-max

Do we have any reason to keep number of frames in GT job strictly less than in the main one?
I can't create GT job with 100% frames, but I can with 100% - 1 frame.

[zhiltsov-max]

@bsekachev,

Do we have any reason to keep number of frames in GT job strictly less than in the main one?
I can't create GT job with 100% frames, but I can with 100% - 1 frame.

This problem is fixed in the PR #6367 (which we still need to merge).

[bsekachev]

Was resolved in #7126