Open4D is developed by the SINRG Lab at Northeastern University as open research infrastructure for time-varying 4D geometry — with the explicit goal of letting research compound instead of resetting with every paper.
4D geometry — meshes and point clouds whose shape, and even connectivity, changes every frame — is the substrate of volumetric video, telepresence, digital twins, and robot perception. Yet the field's tooling is fragmented: every paper rebuilds IO, metrics, and pipelines from scratch.
Open4D exists to fix that. A stable core provides the data model, IO, and metrics; research modules — codecs, trackers, editors — plug into it and evolve independently; benchmarks make comparisons reproducible; and apps prove the whole stack end-to-end, all the way to real-time playback on an XR headset.
The Spatial Intelligence Research Group (SINRG) at Northeastern University works at the intersection of mobile systems, networking, and immersive computing — including volumetric video streaming, XR systems, and machine perception.
SINRG Lab on GitHub →Prof. Mallesham Dasari — Northeastern University. Open4D grows out of a line of research on volumetric video compression and streaming for immersive systems.
m.dasari@northeastern.edu →The .o4d data model with readers, writers, playback, and conversion tools for meshes and point clouds — pip-installable.
N4MC, TSMC, and TVMC integrated; Unity decoder plugin running on Meta Quest 3.
Paper-reproducible evaluation harness with reimplemented baselines and shared metrics.
Performance-critical implementations with Python bindings for real-time pipelines.
Adaptive 4D streaming pipelines and reference applications for telepresence and robotics.
Open4D is an open research platform — contributions are welcome in many forms.
Codecs, trackers, editors, or any algorithm for time-varying geometry, built on the core API.
Reimplemented baselines, new test content, and evaluation configurations.
Quality, rate, latency, and temporal-stability measures for systems-aware evaluation.
Tutorials, examples, and API documentation that lower the barrier to entry.
Optimizations, C++ ports, and profiling for real-time and on-device pipelines.
End-to-end demos — telepresence, robotics, digital twins — that exercise the full stack.
Start with an issue or pull request on GitHub, or reach out directly if you'd like to discuss a larger contribution.
If you use Open4D in your research, please cite it.
@software{open4d2026,
title = {Open4D: An Open Platform for Time-Varying 4D Geometry Data},
author = {Spatial Intelligence Research Group},
year = {2026},
url = {https://github.com/open4dfoundation/Open4D}
}
A CITATION.cff file
with the canonical citation metadata ships in the repository root.
We're building Open4D in the open and would love to hear from researchers and practitioners working with 4D data.