Keypoint MoSeq 0.6.7

Pinning other parts of jax to overwrite Colab's default verisons. 4fef46e

Keypoint MoSeq 0.6.6

Supporting Python 3.12

What's Changed

• Pinned jax<0.7.0. Otherwise, the version of tensorflow_probability installed with Python 3.12 does not use correct JAX syntax. 966ac59
• Also suppressed harmless syntax warnings thrown by Python 3.12's stricter syntax warning checks.

Keypoint MoSeq 0.6.5

Changed generate_syllable_mapping to count frequency by instances rather than frames, which makes it consistent with plot_syllable_frequencies and reindex_syllables_in_checkpoint.

What's Changed

• changed generate_syllable_mapping to count frequency by instances rather than frames, which makes it consistent with plot_syllable_frequencies and reindex_syllables_in_checkpoint 809c404

Keypoint MoSeq 0.6.4

Refactored outlier removal to be a one-liner and added this step to the 'apply model' section of the modeling notebook as well.

What's Changed

• Outlier removal is a one-liner b39b8f3
• Colab notebook mismatches brought up to date with modeling.ipynb 873bc73

Keypoint MoSeq 0.6.3

Added an outlier removal step to the modeling notebook to handle large keypoint jumps. Also added a step to set the sigmasq_loc hyperparameter based on the user's dataset rather than having it hard-coded.

What's Changed

• Keypoint QA (Outlier Removal) by @JRJacoby in #233
• Sigamsq loc hyperparam by @JRJacoby in #235

Full Changelog: 0.6.2...0.6.3

Keypoint MoSeq 0.6.2

• Bug fix: Properly allow overwriting of results_dict in compute_moseq_df

Keypoint MoSeq 0.6.1

• Bug fix: Different length segments were being generated for generate_trajectory_plots, leading to errors.

Keypoint MoSeq 0.6.0

New Additions

• Brand new calibration widget that relies on fewer dependencies and is therefore more stable.
• Added functions for manually merging similar syllables after modeling, described here

Updates

• Fixed bug where the mice would not be properly aligned in side-view when generating grid movies of 3D keypoints.
• More detailed memory requirements in documentation here
• 'pre' and 'post' parameters for syllable visualization functions are now defined in seconds rather than frames. FPS for a dataset is now defined in the config file.
• Batch size automatically determined for input data rather than being hard-coded to maximize memory efficiency by minimizing unnecessary zero-padding.
• Noise added to data during preprocessing is now deterministic for easier debugging.
• Only open a file handle to one video at a time when generating grid movies to avoid max open file handle errors.

Keypoint MoSeq 0.5.2

• Same as 0.5.1, but with a few errors corrected.

Keypoint MoSeq 0.5.1

• Refactored installation to remove backwards compatibility with CUDA11 - CUDA12 is now the only option.
• Removed conda environment yaml files - installation should now happen only with pip. Installation instructions here: https://github.com/dattalab/keypoint-moseq/blob/main/docs/source/install.rst
• Keypoint-moseq now requires python 3.10 or greater instead of python 3.9 or greater.

What's Changed

• setup.cfg updated to match jax-moseq's refactored installation by @JRJacoby in #200
• Installation documentation was updated to match the new setup.cfg by @JRJacoby in #200