X-VLA Soft Fold: 1,500 Robotic Cloth-Folding Episodes

Facebear's dataset contains approximately 1,500 episodes of robotic cloth folding, collected using an Agilex Aloha robotic arm. It was created for the X-VLA paper and demonstrates a near-perfect success rate in the folding task. The dataset was last updated on Hugging Face in November 2025.

Use Cases

• Training imitation learning models for cloth manipulation based on the recorded robotic episodes.
• Benchmarking robotic control policies for folding tasks based on the reported near 100% success rate.
• Developing computer vision models for state estimation of deformable objects like cloth.
• Researching sim-to-real transfer for robotic tasks involving soft materials.

Strengths

• Approximately 1,500 recorded episodes provide a substantial number of task demonstrations.
• Reported near 100% success rate indicates high-performance source data for the specific task.
• Clear association with a published research paper (X-VLA) provides academic context.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.
• Description metadata is limited; actual data quality and content require manual inspection after download.

Provenance

Source

Facebear on Hugging Face, likely associated with academic research (X-VLA paper).

Collection Method

Collected using an Agilex Aloha robotic arm for automated cloth folding.

Freshness

Last updated 2025-11-03 11:54:03; freshness should be verified.