Open4D is an open, research-driven platform for representing, compressing, processing, evaluating, and streaming time-varying 4D geometry data — the stuff that powers real-time dynamic scenes in XR, robotics, teleoperation, digital twins, and autonomy.

Right now, the 4D research ecosystem is fragmented: every paper ships its own objects, IO, codecs, metrics, and pipelines. Open4D gives you:

Stable 4D data models & IO — canonical geometry + temporal primitives.

Modular research modules — compression, tracking, editing.

Benchmarks & metrics — reproducible quality, bitrate, temporal stability evaluations.

Bindings & demos — Python, C#, C++ APIs, end-to-end apps.

Open4D aims to stop reinventing the wheel every time you want to compare methods, build pipelines, or integrate 4D into systems with tight latency/bandwidth constraints.

Repo has docs, examples, and contributions are welcome — especially new modules, datasets/benchmarks, metrics, and real-world pipelines.