v3.2.1

Changes

bugfix

• fix tuple-based group ordering in umap plot
• fix bug with bad initial guess for ellipse fitting

feature

• adds coarse-graining and multiscale bfa
• new convenience wrappers for moving through processing steps from Tracking to Features to Summary
• improved block-smoothing of categorical features
• adds animation functionality to Tracking and Features

docs

• docstrings, getting started and oft_example updates

v3.1.0

• cast boolean results or columns to 'states'

• adds .stored_info() helper for collection objects

• series results in Summary can now also be output to df, and other API improvements

• allows calculation of scalar Summary results by state (e.g. total_distance by cluster)

• plot multiple metrics with SummaryCollection.sns* methods

• specify all_states for state-related Summary metrics

v3.0.0

Changes

• boundaries can now be defined and attached to Features objects for ease of retrieval

• added option to calculate onsets of events of feature data in summary

• added Tracking.copy() and TrackingCollection.copy() commands to create exact copies

• sort point names alphabetically in Tracking.get_point_names()

• BehaviourFlow analysis and seaborn plotting wrappers now accept random seed for reproducibility

• improved docs for the updated API

• more binary operations on BatchResult objects

• chord diagrams can now be plotted from stored transition matrices

• improved clustering for large datasets using streaming approach

• allows hue order to be set independently of display order in sns wrappers

• single Features objects can also be clustered using cluster_embedding()

• merge collections (grouped or flat) into a single flat collection

• allows tracking and features objects (and their collections) to be concatenated (e.g. for real-time chunk/stream processing)

• improved automatic plotting for SummaryCollection object results

• allows nan policies in predictor classes

• example notebooks with mini datasets now available in repo (not yet linked to docs)

• FeaturesCollection.cluster_embedding now tolerates nan values in embedding and still returns a cluster assignment

• Added getters and setters for Tracking objects

• example notebooks with mini datasets now available in repo (not yet linked to docs)

• added version fencing for more robust install

Contributors

@lort3R, @lukasvonziegler, @roessler-f and Lukas von Ziegler

v2.5.0

Changes

✨ Features

• allows non-inplace manipulation of Tracking objects

• add savitzky golay filtering

• improved options for saving figures

🐛 Bug fixes

• smoothing always uses centred rolling window

• Improves data saving to tolerate NaN outputs

• improved nan tolerance in distance calibration

• Fixed error thrown by remove_block() when dealing with pandas series

📚 Docs

• simplified/improved pipeline example generality

@lort3R and @roessler-f

v2.4.0

Changes

✨ Features

• chord diagram plotting for BehaviourFlow

v2.3.0

Changes

✨ Features

• adds BFA histogram plotting and transition matrix UMAP plotting

🐛 Bug fixes

• fix rescaling issue in lowmem clustering

📚 Docs

• Added example to docs for umap plotting function to make input parameter "group" more clear

Contributors

@lort3R and @roessler-f

v2.2.0

Changes

✨ Features

• use tag_info() to get a summary of all the tags attached to a collection
• Adding aggregation functions to summary class
• column helper functions added to Summary class for sum, mean, median, max and min batch processing

🐛 Bug fixes

• improved handling of NaN values in boolean feature results

📚 Docs

• Using Int64 type in EPM pipeline example to handle Nan entries

Contributors

@lort3R and @roessler-f

v2.1.3

Changes

✨ Features

• Tracking.trim() now allows negative endframe values, e.g. endframe = -200 means trim 200 frames from the end

🧹 Chores

• improves compatibility of yolo3r format

@lort3R and @roessler-f

v2.1.2

Changes

📚 Docs

• Simplified README.md to point directly to online docs

Contributors

@lort3R

v2.1.1

Changes

📚 Docs

• update windows and macOS install paths

Contributors

@lort3R