Hello!
First of all, thank you for developing and open-sourcing planTF — it's a great contribution to the planning research community!
I'm currently setting up the environment to experiment with planTF using the nuPlan dataset. However, I noticed that the full nuPlan v1.1 dataset is quite large (several terabytes), which makes downloading and storing the entire dataset challenging, especially on standard research servers.
To better manage storage and bandwidth, I’d like to ask:
Which parts of the nuPlan dataset are strictly required to run training or evaluation with planTF?
Is it sufficient to use only the mini split for initial testing?
Are the raw sensor blobs (e.g., camera images, LiDAR point clouds) necessary, or does planTF rely primarily on the pre-extracted features or scenario DB files?
I’ve reviewed the official nuPlan documentation and the planTF README, but I couldn’t find explicit guidance on minimal dataset requirements.
Any clarification would be greatly appreciated! If there’s a recommended subset or a lightweight setup for getting started, that would be very helpful.
Hello!
First of all, thank you for developing and open-sourcing planTF — it's a great contribution to the planning research community!
I'm currently setting up the environment to experiment with planTF using the nuPlan dataset. However, I noticed that the full nuPlan v1.1 dataset is quite large (several terabytes), which makes downloading and storing the entire dataset challenging, especially on standard research servers.
To better manage storage and bandwidth, I’d like to ask:
Which parts of the nuPlan dataset are strictly required to run training or evaluation with planTF?
Is it sufficient to use only the mini split for initial testing?
Are the raw sensor blobs (e.g., camera images, LiDAR point clouds) necessary, or does planTF rely primarily on the pre-extracted features or scenario DB files?
I’ve reviewed the official nuPlan documentation and the planTF README, but I couldn’t find explicit guidance on minimal dataset requirements.
Any clarification would be greatly appreciated! If there’s a recommended subset or a lightweight setup for getting started, that would be very helpful.