MotIF-1K: Multimodal Human and Robot Motion Trajectories with Task Annotations

MotIF-1K pairs 1,000 multimodal trajectories of human and Stretch-robot motion with task and motion annotations. The dataset was released by authors from MIT and Stanford with the paper 'MotIF: Motion Instruction Fine-tuning' in 2024. It is hosted on Hugging Face by the user 'myconnects'.

Use Cases

• Fine-tuning robot instruction-following models based on the paired motion trajectories and task annotations.
• Researching human-to-robot motion transfer based on the synchronized human and robot motion data.
• Developing multimodal AI systems for robotics based on the dataset's combination of motion and language annotations.

Strengths

• Dataset is explicitly paired, containing synchronized human and robot motion trajectories.
• Includes task and motion annotations, providing structured labels for model training.
• Released with a peer-reviewed research paper (arXiv:2409.10683), suggesting academic rigor.

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 requires manual inspection after download.

Provenance

Source

Released with the paper 'MotIF: Motion Instruction Fine-tuning' by authors from MIT and Stanford.

Collection Method

Likely collected through experiments involving human demonstrations and a Stretch robot.

Time Range

null

Freshness

Last updated 2026-04-18 04:11:36; freshness should be verified.

Geography

null