ETHrobotlearning/tv-task2-clean-aug1_illumination-fixed

tv-task2-clean-aug1_illumination

LeRobot v3 dataset for banana pick-and-place with a SO-101 robot arm.
Built for Vision-Language-Action (VLA / SmolVLA) training.

Dataset summary

Property	Value
Episodes	1 184
Frames	139 368
FPS	10
Task prompts	33
Robot	SO-101 (so_follower)
Camera	Front (480 × 640, AV1)
Action space	6-DoF joint positions
Format	LeRobot v3.0

Construction

Base: ETHrobotlearning/tv-task2-clean-aug1 (592 episodes, trimmed of idle boundary frames).
Each episode is paired with one illumination-augmented copy (brightness ~ Uniform(0.7, 1.2), seed 42), doubling the dataset to 1 184 episodes.

Task prompts (33)

The dataset covers four referring styles:

Color (direct)

• Put the banana in the red / green / blue bowl.
• Put the banana in the red / green / blue colored bowl.
• Place the banana in the red / green / blue bowl.
• Move the banana to the red / green / blue colored bowl.

Color (negation)

• Put the banana into the bowl that is not red and not blue.
• Put the banana into the bowl that is not red and not green.
• Put the banana into the bowl that is not green and not blue.

Ordinal (robot perspective)

• Put the banana into the 1st / 2nd / 3rd bowl from the left from the robot perspective.
• Put the banana into the left / center / right bowl from the robot perspective.
• Put the banana into the leftmost / rightmost / middle bowl from the robot perspective.

Relative position (robot perspective)

• Put the banana into the bowl on the left / right of the red bowl from the robot perspective.
• Put the banana into the bowl on the left / right of the green bowl from the robot perspective.
• Put the banana into the bowl on the left / right of the blue bowl from the robot perspective.
• Put the banana into the bowl between the blue bowl and the green bowl.
• Put the banana into the bowl between the blue bowl and the red bowl.
• Put the banana into the bowl between the red bowl and the green bowl.

Features

Feature	Type	Shape
action	float32	(6,) — joint positions
observation.state	float32	(6,) — joint positions
observation.images.front	video (AV1)	480 × 640 × 3
timestamp	float32	(1,)
frame_index	int64	(1,)
episode_index	int64	(1,)
index	int64	(1,)
task_index	int64	(1,)

Visualization

Browse episodes interactively using the LeRobot visualizer:

For local exploration with tracelr:

huggingface-cli download --repo-type dataset ETHrobotlearning/tv-task2-clean-aug1_illumination --local-dir ./dataset
cargo run --profile opt-dev -- ./dataset

SmolVLA compatibility

lerobot-doctor gate ETHrobotlearning/tv-task2-clean-aug1_illumination --policy smolvla