Endoscopy Images and Text for Visual Question Answering Benchmark

EndoVQA-Instruct is a multi-modal dataset containing endoscopy images and associated text, designed for benchmarking multi-modal large language models in medical analysis. The dataset includes images from the in-house WCE2025 collection and is managed by author Saint-lsy. Access to the data is restricted and requires formal request and approval.

Use Cases

• Benchmark multi-modal LLMs on endoscopy image analysis using the provided image-text pairs for visual question answering tasks.
• Train models to answer medical questions about endoscopy images based on the textual instructions and annotations in the dataset.
• Evaluate the performance of vision-language models on the specific domain of endoscopy, as defined by the EndoBench evaluation framework.

Strengths

• Dataset is the official implementation for the EndoBench paper, providing a standardized benchmark.
• Focuses on the specialized and clinically relevant domain of endoscopy analysis.
• Includes data from the in-house WCE2025 collection, indicating curated medical imagery.

Limitations

• Dataset size, row count, column structure, and file formats are unknown, limiting reproducibility assessments.
• Access is restricted and requires a request and approval process, creating a barrier to immediate use.
• The absence of sample data or column descriptions prevents detailed understanding of the data schema.

Provenance

Source

Saint-lsy on Hugging Face, associated with the EndoBench paper.

Collection Method

Includes images from the in-house WCE2025 dataset; full collection method is not specified.

Freshness

Last updated on 2025-11-01, indicating recent maintenance.