Skip to content

SLEAP v1.1.5

Compare
Choose a tag to compare
@talmo talmo released this 02 Aug 23:34
· 341 commits to main since this release
d31000e

Release of SLEAP v1.1.5.

Includes some features, bug fixes and GUI enhancements.

Highlights

  • New high level APIs:

    • sleap.load_metrics() to load accuracy metrics for a trained model. See example notebook.
    • Labels.copy() to create a deep copy of a SLEAP labels file.
    • Labels.split() to generate random split of the data.
    • Labels.merge_nodes() to merge one node into another (see below).
  • Skeleton editor now supports merging nodes by renaming one with the same name as another. Merging labels with different skeletons is now supported.

Full changelog

  • Empty instance fixes (#569): Some components of the training pipeline would break when there were frames with "empty" instances, i.e., instances with no visible nodes. These are now handled appropriately and removed automatically on pipeline creation as well as DLC data import.

    • Fixes #555
    • Adds Labels.copy() utility function.
  • Small GUI fixes (#570)

    • Add background box to missing instance nodes for better visibility (#565)
    • Add option to change node label font size from menu
    • Increase max instances to be culled to in tracking editor form
    • Fix bug introduced in #562: predicted instances sometimes fail to be colored correctly
  • Fix skeleton merge conflicts (#572)

    • Fix: When merging, union of nodes is now used instead of adding multiple skeletons if node names aren't identical.
    • Feature: Add warning message in merge dialog if nodes are not matching.
    • Feature: Add Labels.merge_nodes() method to support merging one node into another with data update. This is equivalent to
      Skeleton.relabel_nodes() but takes care of updating the point data for the base node.
    • Feature: Enable node merging when renaming nodes from skeleton editor.
  • Add a utility function (Labels.split()) for generating splits (#574)

  • Add high level metrics loader (#573)

  • Fix single instance predictor not using the peak threshold when specified (#571)

Installing

We recommend using Miniconda to install and manage your Python environments. This will also make GPU support work transparently without installing additional dependencies.

See the Installation page in the docs for more info.

Using Conda (Windows/Linux)

  1. Delete any existing environment and start fresh (recommended):
conda env remove -n sleap
  1. Create new environment sleap (recommended):
conda create -n sleap -c sleap sleap=1.1.5

Or to update inside an existing environment:

conda install -c sleap sleap=1.1.5

Using PyPI (Windows/Linux/Mac)

  1. Create a new conda environment (recommended):
conda create -n sleap python=3.6
conda activate sleap
  1. Install from PyPI:
pip install sleap==1.1.5

Or to upgrade an existing installation:

pip install --upgrade --force-reinstall sleap==1.1.5