How can I contribute?
Hey,
Thanks for the amazing dataset. I’ve been exploring how to build some open infrastructure to make it easier to use for both research and production. Some ideas I’ve been thinking about:
- Standardized data loaders and converters (PyTorch, JAX, ROS, etc.)
- Scenario-level access tools for downstream perception, planning, and simulation tasks
- Evaluation + visualization tooling to benchmark models directly on PhysicalAI
Having worked on AV perception at Nissan, I’ve seen how tricky it can be to make large, heterogeneous datasets easy to plug into real pipelines — from preprocessing and synchronization to consistent labeling formats. That’s the pain point I’d like to help solve here.
Before I go too deep, I’d love to hear from the team (and the community):
-> What kind of tooling or wrappers would be most useful to you right now?
-> Are there parts of the dataset ecosystem you’re not actively building out yet, where open contributions might be helpful?
Would love to align on what could be most valuable to develop next.