Google Summer of Code 2026#
GSoC NIU Projects 2026#
NIU is offering a variety of projects for GSoC 2026, organized under different software tools. Click on a project below learn more about the scope and requirements for each one.
A project can be one of three sizes: small (90 h), medium (175 h) or large (350 h). The standard coding period is 12 weeks for medium and large projects, and 8 weeks for small projects.
However, GSoC contributors can request in their proposal up to a 22-week coding period, if they know they may have other commitments or certain weeks when they will not be able to work full time on their GSoC project. During the project preparation period (called “community bonding period”), both the GSoC contributor and the mentors will agree on a schedule and sign off on it. For more details please see our main GSoC page. Please also see our application guidelines and our policy on the use of AI.
Our GSoC ideas are based within specific, larger open source packages we develop. Some of these have specific project ideas associated with them. Others do not yet have specific project ideas, but all on this list welcome ideas developed by GSoC participants. Please reach out to us via our GSoc Zulip channel to discuss.
BrainGlobe#
BrainGlobe is a community-driven suite of open-source Python tools. The BrainGlobe tools are widely used to process, analyse and visualise images of brains (and other related data) in neuroscientific research.
Our working language is English, but our mentors for these projects also speak Italian, French, and German.
Three BrainGlobe repositories (morphapi, brainrender, brainreg) are in pure maintenance mode, and therefore excluded from any GSoC project proposals. All other BrainGlobe repositories welcome alternative project ideas!
Refactor brainglobe-heatmap to use atlas annotations rather than mesh slices for visualisation
brainglobe-heatmap is BrainGlobe’s tool to generate heatmap plots of atlases in 2 and 3D. It relies heavily on BrainGlobe’s meshes, which can cause small imprecisions in visualising data. In 2D it also can fail to slice the meshes along a plane correctly. This could be improved by moving the heatmap functions to relies on the atlas annotations instead of the meshes. There are a number of additional refactoring improvements that could be done to brainglobe-heatmap.
Deliverables
Refactor 2D plotting functionality to use atlas annotations instead of meshes
Tests for new functionality
Ensure any refactored functionality has docstrings
A blog on the BrainGlobe website about the work done
(Stretch goal) - further improvements to
brainglobe-heatmaps
Duration Small (~90 hours)
Difficulty
This project is well suited for an intermediate contributor to open source.
Required skills
Fluency with Python
Nice-to-haves
Experience with pytest.
Experience with visualisation libraries, in particular matplotlib and/or vedo
Experience with neuroanatomy
Potential mentors
Further reading
brainglobe-heatmap issue #103 nicely demonstrates the problem and a potential solution approach.
Improving the user experience in brainrender-napari
brainrender-napari allows BrainGlobe users to download and visualise a variety of neuroanatomical atlases through a graphical user interface. As BrainGlobe supports more and more atlases, we’d like to make it more convenient for users to find the atlas they are interested in, and visualise it in more custom ways.
Deliverables
Add sorting functionality to
brainrender-naparitablesAllow users to filter atlas tables by species
Add functionality to visualise the atlas annotation with preset colours
Any added functionality will require extensive tests and documentation
A blog on the BrainGlobe website about the work done
(Stretch) add functionality to allow users to set the colours of meshes
Duration Large (~350 hours)
Difficulty
This project is well suited for an intermediate contributor to open source.
Required skills
Fluency with Python
Nice-to-haves
Experience with pytest.
Experience with graphical user face libraries, in particular napari and/or Qt
Potential mentors
Further reading
Further details can be found in
brainrender-napariissues #22, #46,#154, and #218Initial familiarisation with
brainrender-napari’s current features may happen by working through the atlas downloading tutorial and the atlas visualisation tutorial
Expand cellfinder to accept different types of input data (several or single channels, 2.5 dimensions)
BrainGlobe’s cellfinder tool allows researchers to detect fluorescent cells in whole-brain microscopy images. It requires whole-brain images of both a signal and a background channels as input, which not many researchers have. We’d like to expand the types of inputs cellfinder support to include brain slices (essentially 2D data) and single channel inputs (no background channel).
Deliverables
Add functionality supporting 2.5-dimensional data
Add functionality to support single-channel data
Any added functionality will require extensive tests and documentation
A blog on the BrainGlobe website about the work done
Duration Large (~350 hours)
Difficulty
This project is well suited for an intermediate contributor to open source.
Required skills
Fluency with Python and NumPy
Nice-to-haves
Experience with pytest.
Experience working with image data
Experience working with large data, e.g. using
pytorchand/ordask
Potential mentors
Further reading
Further details can be found in
cellfinderissues #298 and #352The current
cellfinderfunctionality can be explored by working through the 3D cell detection tutorial
Movement#
Markerless pose estimation tools based on deep learning, such as DeepLabCut and SLEAP, have revolutionised the study of animal behaviour. However, there is currently no user-friendly, general-purpose approach for processing and analysing the trajectories generated by these popular tools. To fill this gap, we’re developing movement, an open-source Python package that provides a unified data analysis interface across pose estimation frameworks.
Our working language is English, but our mentors for these projects also speak Spanish and Mandarin.
Adding I/O support for formats used in human motion tracking
The movement package currently supports a variety of pose estimation formats commonly used in animal behaviour research. With this project, we would like to expand our support to file formats more commonly used in human pose estimation, such as:
Deliverables
Functionality to load at least 3 popular file formats for human motion tracking, such as MMPose output files, COCO keypoint data, motion capture formats or motionBIDS.
Tests to cover any added functionality.
Documentation for the new functionality.
Example use-cases in the
movementgallery.
Duration
Large (~350 hours)
Difficulty
This project is well suited for an intermediate contributor to open source.
Required skills
Fluency with Python, NumPy and/or pandas.
Nice-to-haves
Familiarity with pose estimation frameworks and their usual workflow.
Familiarity with any of the human pose estimation formats mentioned above.
Potential mentors
Further reading
Adding I/O support for popular animal tracking software
The movement package currently supports a variety of pose estimation formats commonly used in animal behaviour research. With this project, we would like to expand our support to other file formats used in the field, such as:
point trackers, such as cotracker3
Bonsai’s centroid tracker
TRex’s pose and bounding box outputs
Deliverables
Functionality to load 2-3 of the file formats mentioned above.
Tests to cover any added functionality.
Documentation for the new functionality.
Example use-cases in the
movementgallery.
Duration
Large (~350 hours)
Difficulty
This project is well suited for an intermediate contributor to open source.
Required skills
Fluency with Python, NumPy and/or pandas.
Nice-to-haves
Familiarity with pose estimation frameworks and their usual workflow.
Familiarity with any of the software packages mentioned above.
Potential mentors
Further reading
Further details are available in the following movement issues: #348, #599
Adding support for tracked segmentation masks
movement has initially focused on pose estimation data, but the long term goal of the package is to support all animal tracking data formats that are relevant to animal behaviour research. With tools like SAM or OCTRON that track segmentation masks in videos, adding support for such data would be a valuable extension of movement’s input capabilities. In this project, we would like to explore how segmentation‑based tracking data can be integrated into the movement framework and design a user-friendly workflow for doing so.
Deliverables
Support for at least one of the file formats output by a popular segmentation and tracking software package.
Tests to cover any added functionality.
Documentation for the new functionality.
Example use-cases in the
movementgallery.
Duration
Large (~350 hours)
Difficulty
This project is well suited for an intermediate contributor to open source.
Required skills
Fluency with Python, NumPy and/or pandas.
Nice-to-haves
Familiarity with segmentation models (such as SAM2, SAM3 or Ultralytics Instance segmentation with YOLO) and their usual workflow.
Familiarity with software packages for instance segmentation applied to biology or animal behaviour research, such as OCTRON or convpaint
Potential mentors
Further reading
Further details are available in the following
movementissue: #301
Adding a module for trajectory complexity metrics
Quantitative characterisation of the trajectories of moving animals is an important component of many behavioural and ecological studies, and also falls within scope for movement. We would like to enable our users to perform statistical characterisation of trajectories through a variety of well-defined metrics such as straightness index, tortuosity, and sinuosity. Functions for computing these metrics could be implemented in a standalone module under movement.kinematics.
Deliverables
Functions for computing at least 3 metrics of trajectory complexity
Tests to cover any added functionality.
Documentation for the new functionality.
Example use-cases in the
movementgallery.
Duration
Medium (~175 hours)
Difficulty
This project is well suited for an intermediate contributor to open source, with a background in research.
Required skills
Nice-to-haves
Research experience in a relevant field: e.g. ethology, neuroscience, behavioural ecology, conservation.
Potential mentors
Further reading
The
trajrR package and its associated paper
Adding a module for metrics of collective behaviour
Understanding how individuals coordinate their movements within a group is a central question in behavioural ecology, ethology, and neuroscience. Since movement already supports multi-individual tracking data, it is well positioned to enable users to quantify collective behaviour through established metrics such as group polarisation (alignment of heading directions), and nearest-neighbour distances. We would like to implement functions for computing collective behaviour metrics in a dedicated module.
Deliverables
A detailed GitHub issue describing suitable metrics to add.
Functions for computing at least 2 metrics of collective behaviour
Tests to cover any added functionality.
Documentation for the new functionality.
Example use-cases in the
movementgallery.
Duration
Medium (~175 hours)
Difficulty
This project is well suited for an intermediate contributor to open source, with a background in research.
Required skills
Nice-to-haves
Research experience in a relevant field: e.g. ethology, neuroscience, behavioural ecology, conservation.
Potential mentors
Further reading
The
swaRmR package and its associated paperAn example script for computing behaviour metrics for a herd of zebras. See also the corresponding preprint.
Ethology#
The main goal of ethology is to facilitate the application of a wide range of computer vision tasks to animal behaviour research, by providing a unified data analysis interface across these tasks.
Our working language is English, but our mentors for these projects also speak Spanish.
Expanding annotations support in ethology
ethology is a package in early development, whose goal is to make it very easy to mix and match computer vision tools to analyse animal behaviour data. Currently, ethology supports loading and curating bounding box annotation datasets.We would like to expand the set of supported annotation types to include mask annotations, and add a UI interface for defining different annotation types in napari.
Deliverables
Support for mask annotation datasets in
ethology(following the current bounding box annotations dataset).Support for defining bounding box annotations in
napari: drawing, exporting to file (e.g. JSON COCO format and asethologydataset netCDF) and loading from file.Support for defining mask annotations in
napari: drawing, exporting to file (e.g. JSON COCO format and asethologydataset netCDF) and loading from file.Example use-cases in the
ethologygallery.(Optional) Support for defining keypoint annotations in
napari
Duration
Large (~350 hours)
Difficulty
This project is well suited for an intermediate contributor to open source.
Required skills
Fluency with Python, experience working with multi-dimensional arrays libraries such as xarray, and experience with pytest.
Nice-to-haves
Experience with napari plugins.
Potential mentors
Further reading
ethologyexamples, to get familiar with the structure of the annotations dataset.
Team#
The NIU GSoC team for 2026 is composed of the following members. To read more about the different roles involved in GSoC, see GSoC Participant Roles.
Our working languages are Python and English ;) - but we also speak other languages! We listed any additional languages spoken by the mentors in the projects list.
