Datasets:
Upload RoboSim scripted training data (real Parquet)
Browse files5 scripted pickup episodes for SO-101, in LeRobot v3 Parquet format
README.md
CHANGED
|
@@ -7,6 +7,7 @@ tags:
|
|
| 7 |
- robosim
|
| 8 |
- pickup
|
| 9 |
- red-cube
|
|
|
|
| 10 |
task_categories:
|
| 11 |
- robotics
|
| 12 |
license: apache-2.0
|
|
@@ -14,7 +15,13 @@ license: apache-2.0
|
|
| 14 |
|
| 15 |
# robosim-red-cube-pickup-test
|
| 16 |
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
## Dataset Details
|
| 20 |
|
|
@@ -27,18 +34,18 @@ RoboSim training data: 5 synthetic "pick up red cube" episodes for the SO-101 ro
|
|
| 27 |
| Total Frames | 450 |
|
| 28 |
| Task | pick_object |
|
| 29 |
| Object | Red Cube |
|
|
|
|
|
|
|
| 30 |
|
| 31 |
-
##
|
| 32 |
|
| 33 |
-
|
| 34 |
-
- **observation.state**: [base, shoulder, elbow, wrist, wristRoll, gripper]
|
| 35 |
-
- **action**: [base, shoulder, elbow, wrist, wristRoll, gripper]
|
| 36 |
|
| 37 |
-
Each episode is a
|
| 38 |
-
1. Approach
|
| 39 |
-
2. Lower to grasp position
|
| 40 |
-
3.
|
| 41 |
-
4. Lift the object
|
| 42 |
|
| 43 |
## Usage with LeRobot
|
| 44 |
|
|
@@ -46,11 +53,15 @@ Each episode is a complete pickup trajectory:
|
|
| 46 |
from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
|
| 47 |
|
| 48 |
dataset = LeRobotDataset("houmans/robosim-red-cube-pickup-test")
|
|
|
|
| 49 |
```
|
| 50 |
|
| 51 |
-
##
|
| 52 |
|
| 53 |
-
|
|
|
|
|
|
|
|
|
|
| 54 |
|
| 55 |
## License
|
| 56 |
|
|
|
|
| 7 |
- robosim
|
| 8 |
- pickup
|
| 9 |
- red-cube
|
| 10 |
+
- simulated
|
| 11 |
task_categories:
|
| 12 |
- robotics
|
| 13 |
license: apache-2.0
|
|
|
|
| 15 |
|
| 16 |
# robosim-red-cube-pickup-test
|
| 17 |
|
| 18 |
+
Synthetic training data: 5 scripted "pick up red cube" episodes for the SO-101 robot arm,
|
| 19 |
+
generated by [RoboSim](https://github.com/hshadab/robosim).
|
| 20 |
+
|
| 21 |
+
> **Note:** This test dataset uses **scripted trajectories** (not MuJoCo physics simulation).
|
| 22 |
+
> The in-app "Generate" button produces higher-quality data using MuJoCo physics + Claude AI
|
| 23 |
+
> for trajectory planning, with domain randomization (varied lighting, camera angles, object
|
| 24 |
+
> positions, and image augmentation).
|
| 25 |
|
| 26 |
## Dataset Details
|
| 27 |
|
|
|
|
| 34 |
| Total Frames | 450 |
|
| 35 |
| Task | pick_object |
|
| 36 |
| Object | Red Cube |
|
| 37 |
+
| Simulation | Scripted (not MuJoCo) |
|
| 38 |
+
| Format | LeRobot v3.0 (real Parquet files) |
|
| 39 |
|
| 40 |
+
## Joint Channels
|
| 41 |
|
| 42 |
+
6-DOF joint observations and actions: `[base, shoulder, elbow, wrist, wristRoll, gripper]`
|
|
|
|
|
|
|
| 43 |
|
| 44 |
+
Each episode is a 3-second scripted pickup trajectory:
|
| 45 |
+
1. **Approach** (0-1s) — arm moves from home to above object
|
| 46 |
+
2. **Lower** (1-1.7s) — arm descends to grasp position
|
| 47 |
+
3. **Grasp** (1.7-2s) — gripper closes
|
| 48 |
+
4. **Lift** (2-3s) — arm lifts the object
|
| 49 |
|
| 50 |
## Usage with LeRobot
|
| 51 |
|
|
|
|
| 53 |
from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
|
| 54 |
|
| 55 |
dataset = LeRobotDataset("houmans/robosim-red-cube-pickup-test")
|
| 56 |
+
print(f"Episodes: {dataset.num_episodes}, Frames: {dataset.num_frames}")
|
| 57 |
```
|
| 58 |
|
| 59 |
+
## How to Generate Better Data
|
| 60 |
|
| 61 |
+
Use RoboSim's in-app training flow for MuJoCo-physics-based data:
|
| 62 |
+
1. Place an object in the 3D scene
|
| 63 |
+
2. Click "Generate" — runs MuJoCo simulation with Claude AI trajectory planning
|
| 64 |
+
3. Click "Upload to HuggingFace" — uploads real Parquet to your HF account
|
| 65 |
|
| 66 |
## License
|
| 67 |
|
data/chunk-000/episode_000000.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:72b1178b424752fda2d0a51b7659affd41b9456b7e306c7bad13ae6019f0355f
|
| 3 |
+
size 9169
|
data/chunk-000/episode_000001.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:16dc75e5db5b1665fce8f5b673e6fade141275dba14a49dc332a4440f8f37dfa
|
| 3 |
+
size 9169
|
data/chunk-000/episode_000002.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:40f484d58b42bcb0c8bcf23489cd838a3e329d502c318bf8f9684b619f0787f3
|
| 3 |
+
size 9169
|
data/chunk-000/episode_000003.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ded05c41fbc9c0bf1f448c1468985f6438b2c8502e0be6a6cec5c8ad96ec48d9
|
| 3 |
+
size 9169
|
data/chunk-000/episode_000004.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4dfb2baff6b5023ee7720ad4d68190ad46106ddf21e54cfa444bb8405e26f448
|
| 3 |
+
size 9169
|
meta/stats.json
CHANGED
|
@@ -1,70 +1,70 @@
|
|
| 1 |
{
|
| 2 |
"observation.state": {
|
| 3 |
"min": [
|
| 4 |
-
-1.
|
| 5 |
-
-45.
|
| 6 |
-
-0.
|
| 7 |
-
-40.
|
| 8 |
-
89.
|
| 9 |
-
19.
|
| 10 |
],
|
| 11 |
"max": [
|
| 12 |
-
1.
|
| 13 |
-
0.
|
| 14 |
-
70.
|
| 15 |
-
0.
|
| 16 |
-
90.
|
| 17 |
-
100.
|
| 18 |
],
|
| 19 |
"mean": [
|
| 20 |
-
0.
|
| 21 |
-
-28.
|
| 22 |
-
51.
|
| 23 |
-
-27.
|
| 24 |
-
89.
|
| 25 |
-
69.
|
| 26 |
],
|
| 27 |
"std": [
|
| 28 |
-
0.
|
| 29 |
-
12.
|
| 30 |
-
19.
|
| 31 |
-
11.
|
| 32 |
-
0.
|
| 33 |
-
37.
|
| 34 |
]
|
| 35 |
},
|
| 36 |
"action": {
|
| 37 |
"min": [
|
| 38 |
-
-1.
|
| 39 |
-
-45.
|
| 40 |
-
-0.
|
| 41 |
-
-40.
|
| 42 |
-
89.
|
| 43 |
-
19.
|
| 44 |
],
|
| 45 |
"max": [
|
| 46 |
-
1.
|
| 47 |
-
0.
|
| 48 |
-
70.
|
| 49 |
-
0.
|
| 50 |
-
90.
|
| 51 |
-
100.
|
| 52 |
],
|
| 53 |
"mean": [
|
| 54 |
-
0.
|
| 55 |
-
-28.
|
| 56 |
-
51.
|
| 57 |
-
-27.
|
| 58 |
-
89.
|
| 59 |
-
69.
|
| 60 |
],
|
| 61 |
"std": [
|
| 62 |
-
0.
|
| 63 |
-
12.
|
| 64 |
-
19.
|
| 65 |
-
11.
|
| 66 |
-
0.
|
| 67 |
-
37.
|
| 68 |
]
|
| 69 |
}
|
| 70 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"observation.state": {
|
| 3 |
"min": [
|
| 4 |
+
-1.2374962541381107,
|
| 5 |
+
-45.24368821393583,
|
| 6 |
+
-0.11422760676412669,
|
| 7 |
+
-40.225808541968235,
|
| 8 |
+
89.7509403802389,
|
| 9 |
+
19.750965589541607
|
| 10 |
],
|
| 11 |
"max": [
|
| 12 |
+
1.2376179291492488,
|
| 13 |
+
0.16322394218873626,
|
| 14 |
+
70.23348025070445,
|
| 15 |
+
0.2116106895886103,
|
| 16 |
+
90.24973660235912,
|
| 17 |
+
100.2490697435544
|
| 18 |
],
|
| 19 |
"mean": [
|
| 20 |
+
0.04567256372482122,
|
| 21 |
+
-28.258325728424246,
|
| 22 |
+
51.946432206710725,
|
| 23 |
+
-27.117487138509063,
|
| 24 |
+
89.99520263003038,
|
| 25 |
+
69.33486891387365
|
| 26 |
],
|
| 27 |
"std": [
|
| 28 |
+
0.6815085279958053,
|
| 29 |
+
12.939659299757693,
|
| 30 |
+
19.609949831082858,
|
| 31 |
+
11.214858381063461,
|
| 32 |
+
0.14369362498120797,
|
| 33 |
+
37.363727868897044
|
| 34 |
]
|
| 35 |
},
|
| 36 |
"action": {
|
| 37 |
"min": [
|
| 38 |
+
-1.3793163982207655,
|
| 39 |
+
-45.238268374708646,
|
| 40 |
+
-0.10205270817114824,
|
| 41 |
+
-40.342223053003686,
|
| 42 |
+
89.62821298567899,
|
| 43 |
+
19.61418778385092
|
| 44 |
],
|
| 45 |
"max": [
|
| 46 |
+
1.3223539374611992,
|
| 47 |
+
0.2867851156067814,
|
| 48 |
+
70.33899806520608,
|
| 49 |
+
0.21077264562037187,
|
| 50 |
+
90.37269254755331,
|
| 51 |
+
100.36870424268649
|
| 52 |
],
|
| 53 |
"mean": [
|
| 54 |
+
0.043622437109475366,
|
| 55 |
+
-28.258169330266508,
|
| 56 |
+
51.954909952305286,
|
| 57 |
+
-27.117939308741114,
|
| 58 |
+
89.99572562057553,
|
| 59 |
+
69.33778023949655
|
| 60 |
],
|
| 61 |
"std": [
|
| 62 |
+
0.678923459433848,
|
| 63 |
+
12.934217351337898,
|
| 64 |
+
19.602608394521123,
|
| 65 |
+
11.2126876844633,
|
| 66 |
+
0.16296718518189104,
|
| 67 |
+
37.36863853934422
|
| 68 |
]
|
| 69 |
}
|
| 70 |
}
|