Question about the csv x/y coordinate of function 'deeplabcut.analyze_videos'

Hi there,
I am now trying to digest every function detail of deeplabcut workflow, and I am a little confused of the x/y coordinate when I use the function ‘deeplabcut.analyze_videos’, because the coordinates generated can have some decimal point, but if the original resolution of the video is 100*100 pixels, then the coordinate of each labeled point should be integer pixels, where do these decimal points come from?
Many thanks!

because of the locref, you can get sub-pixel accuracy/labeling.

Thanks very much for your reply!
Could you please explain a bit more about the locref, or where can I find the definition of locref in the deeplabcut toolbox? Thanks again.

Dear Mathis,
I have found the locref in deeplabcut/pose_estimation_tensorflow folder through github search bar, and I would try to figure it out, thanks.

1 Like