SO-101 Pick and Place Dataset (OpenPi Format)
This dataset contains 40 episodes of a simulated SO-101 robot performing pick-and-place tasks, converted to OpenPi/RLDS format for use with Physical Intelligence's Pi0/Pi0.5 models.
Source
Converted from LeRobot dataset: danbhf/sim_pick_place_merged_40ep
Format
Each episode is stored as an NPZ file containing:
| Key | Shape | Type | Description |
|---|---|---|---|
observation/state |
(N, 6) | float32 | Joint positions (6 DoF) |
observation/image |
(N, 480, 640, 3) | uint8 | Overhead camera RGB |
observation/wrist_image |
(N, 480, 640, 3) | uint8 | Wrist camera RGB |
action |
(N, 6) | float32 | Joint actions (6 DoF) |
language_instruction |
string | - | Task description |
is_first |
(N,) | bool | First step in episode |
is_last |
(N,) | bool | Last step in episode |
is_terminal |
(N,) | bool | Terminal state |
Statistics
- Episodes: 40
- Total Steps: 6,559
- Task: "Pick up the block and place it in the bowl"
- Robot: SO-101 (6 DoF + gripper)
- Cameras: Overhead (main), Wrist
Usage
import numpy as np
# Load an episode
data = np.load("episode_000000.npz", allow_pickle=True)
# Access observations
states = data["observation/state"] # (N, 6)
images = data["observation/image"] # (N, H, W, 3)
actions = data["action"] # (N, 6)
language = str(data["language_instruction"])
License
Same as source dataset.