Release v0.0.21 and next steps#

This is our inaugaural blogpost, containing a summary of the v0.0.21 release and a preview of what’s coming next in 2025.

What’s new in movement v0.0.21?#

Tip

See our installation guide for instructions on how to install the latest version or upgrade from an existing installation.

Input/Output

We have added the movement.io.load_poses.from_multiview_files() function to support loading pose tracking data from multiple camera views.
We have made several small improvements to reading bounding boxes tracks. See our new example to learn more about working with bounding boxes.
We have added a new example on using movement to convert pose tracking data between different file formats.

Kinematics

The kinematics module has been moved from movement.analysis.kinematics to movement.kinematics and packs a number of new functions:

Breaking changes

We have dropped support for using filtering and kinematic functions via the move accessor syntax, because we’ve found the concept hard to convey to new users. All functions are henceforth solely accessible by importing them from the relevant modules. Having one way of doing things simplifies the mental model for users and reduces the maintenance effort on our side. See an example below:
```
# Instead of:
position_filt = ds.move.median_filter(window=5)
velocity = ds.move.compute_velocity()

# Use:
from movement.filtering import median_filter
from movement.kinematics import compute_velocity

position_filt = median_filter(ds.position, window=5)
velocity = compute_velocity(ds.position)
```

We have slightly modified the structure of movement datasets, by changing the order of dimensions. This should have no effect when indexing data by dimension names, i.e. using the xarray.Dataset.sel() or xarray.Dataset.isel() methods. However, you may need to update your code if you are using Numpy-style indexing, for example:

# Indexing with dimension names (recommended, works always)
position = ds.position.isel(
    individuals=0, keypoints=-1     # first individual, last keypoint
)

# Numpy-style indexing with the old dimension order (will no longer work)
position = ds.position[:, 0, -1, :]  # time, individuals, keypoints, space

# Numpy-style indexing with the updated dimension order (use this instead)
position = ds.position[:, :, -1, 0]  # time, space, keypoints, individuals

Looking to v0.1 and beyond#

Over the last 1.5 years, we have gradually built up the core functionalities we envisioned for movement version v0.1, as described in our roadmap. These have included input/output support for a few popular animal tracking frameworks, as well as methods for data cleaning and computing kinematic variables.

What we’re still missing is a napari plugin for movement, which we envision both as an interactive visualisation framework for motion tracking data as well as a graphical user interface for movement. We have been working on a minimal version of this plugin for a while and are expecting to ship it as part of the v0.1 release in early 2025.

After v0.1, we’ll be switching to semantic versioning, as it applies to MINOR (new features) and PATCH (bug fixes) versions. Until we are ready for a v1 MAJOR version, we cannot commit to backward compatibility, but any breaking changes will be clearly communicated in the release notes.

Announcing movement Community Calls#

We are committed to fostering openness, transparency, and a strong sense of community within the movement project. Starting next year, we will host regular Community Calls via Zoom.

The calls will take place every second Friday from 11:00 to 11:45 GMT, beginning on 10 January 2025. These calls are open to anyone interested in contributing to movement or sharing feedback on the project’s progress and direction.

A few days before each call, we will post an announcement on Zulip with the Zoom link and agenda. We encourage everyone who’s interested in joining to follow this Zulip topic to stay updated.