sgoldenlab / simba

SimBA (Simple Behavioral Analysis), a pipeline and GUI for developing supervised behavioral classifiers
https://simba-uw-tf-dev.readthedocs.io/
GNU General Public License v3.0
272 stars 137 forks source link

Identify Behavior Screen Opens as Blank #320

Open yampants opened 5 months ago

yampants commented 5 months ago

Describe the bug I created a SimBa project and added one video alongside the slp file with the same name which included the pose estimation. When I went to the Label behavior tab and tried to annotate a new video, a new tab opened which said SimBa at the top but was completely blank. I tried repeatedly with new videos to get the screen to open, but it is always blank.

To Reproduce Steps to reproduce the behavior:

  1. Create new SimBa Project with one video and slp data
  2. Open project in SimBa
  3. Click on Label Behavior
  4. Click on 'Select video (create new video annotation)'
  5. See error

Expected behavior I believe it should be a screen similar to the one that opened when I identified the two animals initially.

Desktop (please complete the following information):

sronilsson commented 5 months ago

Hi @yampants and thanks for reporting. One question to help me find the issue: When the blank screen pops open, do you see any error msgs in the main Windows terminal window you used to open SimBA with? If so, would you mind sending me a screengrab of the error?

yampants commented 5 months ago

Hi @sronilsson, of course! I went to the file location in the error and the video does exist there, but I'm not quite sure I understand the error. Screenshot 2024-01-25 143740

sronilsson commented 5 months ago

Ah got thanks @yampants! When labelling videos, SimBA will need a file to append the annotations to. This file lives in the project_folder/csv/features_extracted, and will have the same name your video file. To create it, try clicking on the Extract features button and let me know how it goes:

image

That said.. just to check: the video name labels.v001 look a little suspicious. Did you feed in a video called labels.v001 to SLEAP or deeplabcut, and sleap/deeplabcut generated a file called labels.v001 containing body-part predictions for the video labels.v001?