ManiTwin-100K: 1,000 Sample Objects for Robotic Manipulation Research

ManiTwin-100K is a large-scale dataset of manipulation-ready digital object twins for robotic manipulation research. It includes approximately 1,000 sample objects, each with simulation-ready 3D meshes, physical properties, functional point annotations, grasp configurations, and language descriptions. The dataset was created by ManiTwin and was last updated on Hugging Face in April 2026.

Use Cases

• Training robotic manipulation models based on the provided grasp configurations and physical properties.
• Developing simulation environments using the simulation-ready 3D meshes and validated physics.
• Researching object affordances and functional understanding based on the functional point annotations.
• Building language-conditioned robotic systems based on the rich language descriptions paired with 3D objects.

Strengths

• Includes approximately 1,000 sample objects, indicating a substantial collection.
• Contains multiple validated data modalities per object, including 3D meshes, physical properties, and annotations.
• Data is validated through physics-based simulation, suggesting a focus on real-world applicability.

Limitations

• The full dataset of 100K objects is not yet released; only a 1K sample is available.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the full dataset is unknown, which may limit suitability assessment.

Provenance

Source

ManiTwin

Collection Method

Created as digital object twins; details of gathering method are not specified.

Freshness

Last updated 2026-04-03 11:41:48; freshness should be verified.