Is there an existing issue for this?

• I have searched the existing issues

Bug description

Upgraded to the most recent DLC release (2.3) and tried to recreate what I used to do in previous versions.

1. Analyze video
2. Open GUI and click on extract outlier frames, choose video and choose "manual" as option
3. Click on "Extract frames":

• Expected: Frame extraction GUI opens up with the video, a slider, and the points overlayed on top
• Reality: A warning (see below) and an empty napari window
  WARNING: QStatusBar::insertPermanentWidget: Index out of range (0), appending widget
  -> Decided to load the video and .h5 file of the analysis instead.

4. Video and points are overlayed.
   -> realized that the points and video are one frame off (I have a flash in the video where no points are detected, which is the correct behavior and in the opened video the "lack of detections" was constantly one frame off).
5. Go to chosen frames and click "extract frame"

• Expectation: Images are stored in labelled_data along with a machine-labels file
• Reality: only extracted images without machine-labels stored in labelled_data

• This requires me to manually label the 50 manually extracted images. Normally no problem, but this model has 60 points to track (expanded monkey face from the Model Zoo). And I guess this is also not the intended behavior.

Operating System

macOS Monterey

DeepLabCut version

2.3

DeepLabCut mode

single animal

Device type

gpu

Steps To Reproduce

No response

Relevant log output

No response

Anything else?

It is possible that this has something to do with the project being created with an older version of DLC (should be around May 2022).

Code of Conduct

• I agree to follow this project's Code of Conduct