Yinpei/robomme_data_h5

RoboMME Training Data (H5 format)

Arxiv Paper | HF Paper | Website | Benchmark Code | Policy Learning Code

H5 format of RoboMME training data. There are a total of 16 tasks, each with 100 episodes.

.
├── README.md
├── record_dataset_BinFill.h5
├── record_dataset_ButtonUnmask.h5
├── record_dataset_ButtonUnmaskSwap.h5
├── record_dataset_InsertPeg.h5
├── record_dataset_MoveCube.h5
├── record_dataset_PatternLock.h5
├── record_dataset_PickHighlight.h5
├── record_dataset_PickXtimes.h5
├── record_dataset_RouteStick.h5
├── record_dataset_StopCube.h5
├── record_dataset_SwingXtimes.h5
├── record_dataset_VideoPlaceButton.h5
├── record_dataset_VideoPlaceOrder.h5
├── record_dataset_VideoRepick.h5
├── record_dataset_VideoUnmask.h5
└── record_dataset_VideoUnmaskSwap.h5

For more information about dataset usage, please refer to https://github.com/RoboMME/robomme_benchmark?tab=readme-ov-file#-training-data

We also provide other formats of RoboMME training data for people to use:

1. Pickle format, which is used in our MME-VLA experiments.
2. Lerobot format, which is used in our Diffusion Policy experiments.
3. TensorFlow Dataset format, which is used in our MemoryVLA experiments.

Download this dataset using either of the following methods:

1. git clone git@hf.co:Yinpei/robomme_data_h5
2. hf download Yinpei/robomme_data_h5
3. from huggingface_hub import snapshot_download; snapshot_download(repo_id="Yinpei/robomme_data_h5", repo_type="model") in python code.