Add dataset card
Browse files
README.md
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pretty_name: Piston Long-Horizon Videos
|
| 3 |
+
tags:
|
| 4 |
+
- robotics
|
| 5 |
+
- simulation
|
| 6 |
+
- video
|
| 7 |
+
- isaaclab
|
| 8 |
+
- replay
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# Piston Long-Horizon Videos
|
| 12 |
+
|
| 13 |
+
This dataset contains 107 long-horizon piston replay recordings generated from the local archive `piston_long_horizon_ego_wrist_videos_20260726.tar`.
|
| 14 |
+
|
| 15 |
+
Each trajectory has two synchronized camera-view videos:
|
| 16 |
+
|
| 17 |
+
- `ego/`: ego/front camera recordings
|
| 18 |
+
- `wrist/`: right wrist camera recordings
|
| 19 |
+
|
| 20 |
+
The videos were rendered from the 50 Hz replay pipeline used for the Piston task. The replay uses a free-physics object mode rather than strictly forcing the object pose at every frame, so the videos are intended for visual observation/replay inspection rather than as a ground-truth physical-state dataset.
|
| 21 |
+
|
| 22 |
+
## Directory Structure
|
| 23 |
+
|
| 24 |
+
```text
|
| 25 |
+
ego/
|
| 26 |
+
trajectory_csv_1/
|
| 27 |
+
episode_trial0_env0.mp4
|
| 28 |
+
...
|
| 29 |
+
trajectory_csv_2/
|
| 30 |
+
...
|
| 31 |
+
trajectory_csv_3/
|
| 32 |
+
...
|
| 33 |
+
wrist/
|
| 34 |
+
trajectory_csv_1/
|
| 35 |
+
episode_trial0_env0.mp4
|
| 36 |
+
...
|
| 37 |
+
trajectory_csv_2/
|
| 38 |
+
...
|
| 39 |
+
trajectory_csv_3/
|
| 40 |
+
...
|
| 41 |
+
manifest.csv
|
| 42 |
+
```
|
| 43 |
+
|
| 44 |
+
## Manifest
|
| 45 |
+
|
| 46 |
+
`manifest.csv` contains one row per trajectory and aligns the two camera views:
|
| 47 |
+
|
| 48 |
+
| Column | Description |
|
| 49 |
+
| --- | --- |
|
| 50 |
+
| `trajectory_group` | Source trajectory group, for example `trajectory_csv_1`. |
|
| 51 |
+
| `episode` | Episode identifier, for example `episode_trial0_env0`. |
|
| 52 |
+
| `ego_video` | Relative path to the ego/front camera video. |
|
| 53 |
+
| `wrist_video` | Relative path to the right wrist camera video. |
|
| 54 |
+
|
| 55 |
+
## Counts
|
| 56 |
+
|
| 57 |
+
- Trajectories: 107
|
| 58 |
+
- Ego/front camera videos: 107
|
| 59 |
+
- Right wrist camera videos: 107
|
| 60 |
+
- Total MP4 files: 214
|
| 61 |
+
|
| 62 |
+
## Loading Example
|
| 63 |
+
|
| 64 |
+
```python
|
| 65 |
+
import pandas as pd
|
| 66 |
+
|
| 67 |
+
manifest = pd.read_csv("manifest.csv")
|
| 68 |
+
row = manifest.iloc[0]
|
| 69 |
+
|
| 70 |
+
ego_path = row["ego_video"]
|
| 71 |
+
wrist_path = row["wrist_video"]
|
| 72 |
+
```
|
| 73 |
+
|
| 74 |
+
## Limitations
|
| 75 |
+
|
| 76 |
+
- This dataset contains rendered replay videos only.
|
| 77 |
+
- It does not include the original CSV trajectories, actions, robot states, or object states.
|
| 78 |
+
- The recordings are simulated IsaacLab/Unitree G1 replay outputs, not real robot camera footage.
|
| 79 |
+
- No train/validation/test split is provided.
|