Issue with analyzing multiple body parts

Have installed dlc 2.1.

I trained the network after labeling videos containing one mouse, and wrote ‘num_outputs: 2’ in the config file. But when I ran the function ‘deeplabcut.analyze_video()’ to analyze new videos containing two mice, an error appeared:

ValueError: could not broadcast input array from shape (8,30) into shape (8,15)

It seemed the error was from line 263 of the file ‘predict_videos.py’.
I don’t know how to handle it. Looking forward to your reply.

@wdy

I believe @AlexanderMathis @MWMathis might be the best persons to answer your question…

1 Like

For now set: TFGPUinference=True (is that set to true?)

1 Like

Thanks for reply @AlexanderMathis .
I tried setting ‘TFGPUinference=True’ when using the function(see below), but the same error still appeared.

deeplabcut.analyze_videos(config_path, videos, shuffle=1, save_as_csv=True, videotype=’.avi’, TFGPUinference=True)

I assume you have 5 body parts? I also assume that you set num_outputs: 2 in the /dlc-models/…/test/pose_cfg.yaml. Can you then please post the output of DLC so that I can see what is happening before the error.

@AlexanderMathis
I uploaded 3 screenshots about the config.yaml, the /dlc-models/…/test/pose_cfg.yaml and the error output.

Ok, sorry! I just checked this and there was a bug. Somehow the DLC2.1 rolled back to an earlier version in that aspect. I will upload DLC 2.1.1 shortly!

Thanks for finding this!

1 Like

It’s out: https://pypi.org/manage/project/deeplabcut/release/2.1.1/
Please give it a try!

Thanks very much. It works!

1 Like

Here is the link: https://github.com/AlexEMG/DeepLabCut/pull/321