PMC-OA: 1.6M Biomedical Image-Caption Pairs from PubMedCentral

1.6 million biomedical image-caption pairs were collected from the PubMedCentral OpenAccess subset to address data scarcity for foundation models. The dataset, released by author axiong in August 2024, is described as 8 times larger than previous collections. It features fine-grained alignment between subfigures and subcaptions, covering diverse modalities and diseases.

Use Cases

• Pretraining CLIP-style vision-language models based on the large-scale image-text pairs.
• Improving biomedical image-text retrieval performance based on the aligned subfigure and subcaption data.
• Enhancing medical visual question answering (VQA) models based on the described multimodal content.
• Benchmarking image classification models on medical tasks based on the dataset's coverage of diverse diseases.

Strengths

• Contains 1.6 million image-caption pairs, which is described as an 8x scale increase over prior collections.
• Features fine-grained alignment at the subfigure and subcaption level for the majority of samples.
• Covers a diverse range of biomedical modalities and diseases as mentioned in the description.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific modeling needs.
• Last updated 2024-08-30 03:29:29; freshness should be verified for time-sensitive applications.

Provenance

Source

PubMedCentral OpenAccess subset.

Collection Method

Collected and released by author axiong.

Freshness

Last updated 2024-08-30 03:29:29.