SurgLaVi: Large-Scale Hierarchical Surgical Video Dataset with Captions

SurgLaVi is a large-scale hierarchical dataset for surgical vision-language representation learning, created by researchers from Intuitive Surgical Inc. and Universidad de los Andes. The dataset is associated with a paper on arXiv (2509.10555) and was last updated on the Hugging Face platform in March 2026.

Use Cases

• Zero-shot classification of surgical scenes based on video and text descriptions
• Contrastive learning for aligning surgical video frames with textual captions
• Video classification for identifying surgical phases or actions
• Training vision-language models for medical procedure understanding

Strengths

• Dataset is described as 'large-scale' and 'hierarchical' in the title and platform tags
• Associated with a formal research paper (arXiv:2509.10555) and code repository
• Platform tags indicate it contains video, captions, and is multimodal

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count, file formats, and exact size are unknown, which may limit suitability assessment
• Data may reflect bias inherent to the specific surgical procedures and sources used

Provenance

Source

Intuitive Surgical Inc. and Universidad de los Andes

Collection Method

Likely collected from surgical procedures; specific methodology is not detailed in the provided input.

Time Range

null

Freshness

Last updated 2026-03-03 23:45:49; freshness should be verified

Geography

Platform tags indicate 'Regionus', but specific geographic coverage is not detailed.