yuanzhangMIT

GEAR-VLA Dataset

Name: GEAR-VLA Dataset
Creator: yuanzhang
License: https://opensource.org/license/mit
Keywords: rgb, language, affordance-grounding, visual-question-answering

Image-language dataset for object affordance-area grounding, where models identify interaction regions as bounding boxes given an image and natural-language affordance query.

Downloads0
Episodes12448
Likes1

Why This Matters for Physical AI

This dataset grounds visual affordances to natural language queries, enabling embodied AI systems to understand where and how to interact with objects in their environment.

Technical Profile

Modalities: rgblanguage
Task Types: affordance-groundingvisual-question-answering
Episodes: 12448
Data Format: Parquet
Annotation Types: language_instructionsbounding_boxes
License: MIT

Part of the GEAR-VLA family

Access

View on HuggingFace Read Paper

Need custom rgb data?

Claru builds purpose-built datasets for any environment applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets

Test_2_20260714_160610

A robotics dataset created using LeRobot containing teleoperated robot arm manipulation data with RGB video and proprioceptive observations.

rgbproprioception

null downloadsJul 2026apache-2.0gated

Robo-MLT: Medical Laboratory Manipulation Datasets

Real-robot teleoperated demonstrations for two long-horizon medical laboratory workflows (blood gas analysis and tube sorting) collected on an Agilex dual-arm Static-Aloha platform with vision-language-action supervision.

rgbproprioceptionlanguage

null downloadsJul 2026apache-2.0gated

dvla-place5obj-roll-250hz-events

A 5-object, 3-receptacle generalization of the can-into-bowl task with rolling object dynamics at 250 Hz, featuring both RGB and DVS event camera data. Contains 10,083 episodes where a robot must catch a rolling object and place it in a named receptacle.

rgbevent_cameraproprioception

null downloadsJul 2026apache-2.0gated

LeRobot LeLab Move Object Dataset

A robotics manipulation dataset created with LeRobot containing 44 episodes of a robot moving objects, captured at 30 fps with multi-camera observations and joint position actions.

rgbproprioception

null downloadsJul 2026apache-2.0gated

Masterji/so_101_plastic_20260803_175809

A robotics dataset created using LeRobot containing 10 episodes of manipulation tasks collected at 30 fps with multi-view RGB video and proprioceptive state observations from a SO follower robot.

rgbproprioception

null downloadsAug 2026apache-2.0

W2-VLA Offline CoT Labels

Frame-aligned offline chain-of-thought annotations for training vision-language-action policies on LIBERO, RoboTwin, and four real-world manipulation tasks. Contains annotations only; images, videos, and actions are not included.

language

null downloadsAug 2026