README.md

---
license: apache-2.0
task_categories:
- robotics
tags:
- robotics
- lerobot
- yam
- teleoperation
- imitation-learning
- manipulation
---

# yam-pick-duster-200 — joint space (LeRobot v3.0)

200 teleoperated demonstrations of a single I2RT YAM arm picking up a blue
duster and placing it in a red box, recorded in VR. **Joint-space** state and
action, two camera views.

An end-effector-space version of the *same 200 takes* is published separately as
[`Dimios45/yam-pick-duster-200-ee`](https://huggingface.co/datasets/Dimios45/yam-pick-duster-200-ee) —
same episodes, same wall-clock spans, different action space and format. Both
were written from one control loop, so episode *N* is the same demonstration in
each.

## At a glance

| | |
| --- | --- |
| Episodes | 200 |
| Frames | 76,307 @ 25 Hz (50.9 min) |
| Episode length | 109–280 frames (10.8–27.9 s), median 14.5 s |
| Robot | I2RT YAM, 6-DoF + `linear_4310` gripper, right arm only |
| Cameras | `top` (fixed overhead), `right_wrist` — both 640×480 RGB, **uncropped** |
| Task | `"pick up the duster"` (pick the duster, place it in the box) |

Recorded in two sittings of 100 on 2026-08-17/18.

## Schema

```
observation.state  float32 (7,)  [right_joint_1..6 (rad), right_gripper]   measured
action             float32 (7,)  same layout                               commanded
observation.images.top          video (480, 640, 3)
observation.images.right_wrist  video (480, 640, 3)
```

**Gripper convention: 0 = open, 1 = closed** — the inverse of i2rt's native
normalisation, converted at record time.

`action` is what the teleoperator commanded on that tick; `observation.state` is
what the arm measured. The command leads the measurement by a few ticks, as
expected of a position-controlled arm under load.

## Data quality

Measured over all 200 episodes:

- **86% of frames contain motion** (range 64–94%). No idle or dead takes.
- **Start pose repeatability: 0.08 / 0.33 / 0.11 mm std** — every episode begins
  from the same home pose (the operator ramps the arm home between takes).
- **Lighting is stable**: top-camera brightness std 2.4/255, with per-50-episode
  medians of 144.4 / 143.8 / 144.0 / 144.1 — no drift across either sitting.
- **198/200 episodes contain a grasp.** Two takes (`20260818T004003534366`,
  `20260818T005359521504`) were aborted without closing the gripper; exclude
  them at conversion time if you want a clean set.
- **14 episodes re-grasp** (two close events) — a failed first attempt followed
  by a retry. Useful recovery behaviour, but worth knowing it is in there.

### Object placement

Both the duster and the box were repositioned between episodes, so neither is
at a fixed location:

| | x range | y range | std |
| --- | --- | --- | --- |
| grasp (duster pickup) | 0.35–0.60 m | −0.15 to +0.25 m | 52 / 119 mm |
| release (into the box) | 0.32–0.60 m | −0.14 to +0.25 m | 63 / 123 mm |

Coverage is not uniform — the first 100 episodes concentrated in two bands with
a gap around y ≈ +0.05, which the second 100 partly filled. Reach spans roughly
50 × 48 × 31 cm in x/y/z.

## Recording setup

Meta Quest controllers → WebXR → differential IK → joint commands, using
[vr-teleop-kit](https://github.com/Dream-Machines-Robotics/vr-teleop-kit).
The operator holds a grip button to clutch the arm; the trigger drives the
gripper.

The arm is commanded at **200 Hz** and the dataset is *sampled* from that loop
at 25 Hz. Rate matters: an earlier version commanded at the dataset rate and the
arm was visibly jittery, because it received a new joint target only every
1/fps s and the IK's per-tick velocity cap tightened by the same factor.

## Loading

```python
from lerobot.datasets.lerobot_dataset import LeRobotDataset

ds = LeRobotDataset("Dimios45/yam-pick-duster-200")
item = ds[0]
item["observation.state"]              # (7,) joints + gripper
item["observation.images.top"]         # (3, 480, 640) float32 in [0, 1], RGB
```

Works directly with LeRobot-native policies (ACT, diffusion policy, pi0,
SmolVLA).

## Training a B-spline diffusion policy on this

The [bspline-policy](https://github.com/haoyu-x/simple_mobile_bsp) stack reads
robomimic HDF5. `to_robomimic.py` is included in this repo:

```bash
python to_robomimic.py --from lerobot \
    --repo-id Dimios45/yam-pick-duster-200 --root <local-root> \
    --output-path yam_joint.hdf5
```

Yields `obs/joint_pos (N,7)`, `obs/top_image`, `obs/wrist_image` (84×84 RGB) and
`actions (N,7)` — the stack's `single_yam_joint` format, which needs no rotation
conversion and **no IK at deployment**. Matching `shape_meta`:

```yaml
shape_meta: &shape_meta
  obs:
    top_image:   {shape: [3, 84, 84], type: rgb}
    wrist_image: {shape: [3, 84, 84], type: rgb}
    joint_pos:   {shape: [7]}
  action:
    shape: [7]
```

With `action.shape: [7]`, `_convert_actions` passes the joint targets through
untouched — that is intended, not a bug.

## Cropping

Frames are stored **uncropped** so the crop can be retuned without re-recording.
The overhead camera's useful region is roughly `x=42, y=28, w=598, h=414`; the
wrist camera sees the room above the table horizon, but that crop is
**pose-dependent** (the horizon moves as the arm pitches), so verify across your
workspace before committing.

Whatever crop you train with must be applied identically at deployment. The
converter stamps the crop into the HDF5 attributes so the choice travels with
the data.

## Licence

Apache-2.0.