Closed anshu957 closed 1 year ago
Although it was linked to the issue, it was not mentioned in this pull request:
Going from shapely 1.7 -> 2.0 seems to change the way of accessing centroid keypoints. This patch should fix it: https://github.com/KumarLabJax/JABS-behavior-classifier/pull/27/commits/188fe7a654d55a4a34b2c689f1f8fe457ea66929
It appears that centroid keypoints are only accessed erroneously in multimouse features (via the social distance angles), so single mouse projects are fine.
Added the 2 patches and bumped the version.
I didn't run extensive tests, but I did run initialize_project.py
on a multimouse project (via the venv libraries) to make sure the patches were compatible.
If @anshu957 wanted to do one last check for this, you can... otherwise I think we can merge/release.
Thanks @SkepticRaven. I just validated it for a single mouse dataset and I could train and export with all three algorithms without any problem. It is displaying the overlaying poses and tracks correctly as well. I think we can merge it and release it.
Description:
This pull request adds support for Mac M1/M2 devices by setting the QT_MAC_WANTS_LAYER environment variable to 1. This is required to ensure proper rendering of the application's graphical interface.Changes:
Added the following code to the app.py:
Added conda environment yaml file for users who want to install using conda as pip has problems installing
pyside2
on Mac M1/M2 (https://forum.qt.io/topic/124289/mac-m1-pyside-2-6-can-t-install-lib/8)Testing:
Tested this on 2 different mac M1 devices and verified that the application's graphical interface is rendered correctly.