Episodes Preview SO-101 Visualizer
32 episodes · 10 fps · 1 camera · 640×480 av1

SO-101 pick-and-place - 16 teleop + 16 handheld episodes

The 16-episode teleoperated base plus 16 handheld demonstrations recorded with a GoPro on a hand-held gripper and retargeted into joint space. This is the substitution condition.

episodes 32
frames 9,376
fps 10
robot type so101_follower
camera wrist, 640x480
action / state 6 absolute joint positions, degrees

16 teleoperated episodes (5,600 frames, 59.7% of frames) and 16 handheld episodes (3,776 frames). The handheld episodes are shorter, so they contribute fewer frames than their episode count suggests - this mixture holds 16% fewer frames than the 32-teleop set it is compared against.

Trained checkpoint: robotfuel/act_so101_t16b_u16. All three training sets in this study are published so the comparison can be rerun exactly; the recipe is identical across them (50,000 steps, batch 8, seed 1000) and the dataset is the only variable.

Load it

from lerobot.datasets.lerobot_dataset import LeRobotDataset

ds = LeRobotDataset("robotfuel/so101_t16b_u16")

Sources

Where this comes from

This repo is one artifact of a single study: can cheap handheld demonstrations substitute for robot teleoperation? The three policies below differ only in training data, so they are only meaningful read together.

Policies

Datasets

Every success rate published anywhere in this set is re-derived from the raw records by the verifier script; no number is transcribed by hand.

Downloads last month
94

Models trained or fine-tuned on robotfuel/so101_t16b_u16