The Dataset Viewer has been disabled on this dataset.

Raw Dataset for the 1X World Model Sammpling Challenge.

Download with:

huggingface-cli download 1x-technologies/worldmodel_raw_data --repo-type dataset --local-dir data

Train/Val v2.0

The training dataset is shareded into 100 independent shards. The definitions are as follows:

• video_{shard}.mp4: Raw video with a resolution of 512x512.
• segment_idx_{shard}.bin - Maps each frame i to its corresponding segment index. You may want to use this to separate non-contiguous frames from different videos (transitions).
• states_{shard}.bin - States arrays (defined below in Index-to-State Mapping) stored in np.float32 format. For frame i, the corresponding state is represented by states_{shard}[i].
• metadata - The metadata.json file provides high-level information about the entire dataset, while metadata_{shard}.json files contain specific details for each shard.

Test v2.0

We provide a 250 sample test_v2.0 dataset for the World Model Compression Challenge with a similar structure (video_{shard}.bin, states_{shard}.bin). Use: - unpack_data_test.py to read the test set - unpack_data_train_val.py to read train/val

Index-to-State Mapping (NEW)

 {
 0: HIP_YAW
 1: HIP_ROLL
 2: HIP_PITCH
 3: KNEE_PITCH
 4: ANKLE_ROLL
 5: ANKLE_PITCH
 6: LEFT_SHOULDER_PITCH
 7: LEFT_SHOULDER_ROLL
 8: LEFT_SHOULDER_YAW
 9: LEFT_ELBOW_PITCH
 10: LEFT_ELBOW_YAW
 11: LEFT_WRIST_PITCH
 12: LEFT_WRIST_ROLL
 13: RIGHT_SHOULDER_PITCH
 14: RIGHT_SHOULDER_ROLL
 15: RIGHT_SHOULDER_YAW
 16: RIGHT_ELBOW_PITCH
 17: RIGHT_ELBOW_YAW
 18: RIGHT_WRIST_PITCH
 19: RIGHT_WRIST_ROLL
 20: NECK_PITCH
 21: Left hand closure state (0 = open, 1 = closed)
 22: Right hand closure state (0 = open, 1 = closed)
 23: Linear Velocity
 24: Angular Velocity
 }