U-RxnDiagram-15k: 15,400 Chemical Reaction Images from Scientific Literature

15,400 annotated chemical reaction diagrams extracted from scientific literature PDFs form this large-scale dataset. It was created by songjhPKU and is designed to support research in cheminformatics, document analysis, and computer vision. The dataset was last updated on March 11, 2026.

Use Cases

• Training computer vision models for chemical reaction diagram recognition based on annotated images.
• Developing multimodal models for document analysis based on images extracted from PDFs.
• Benchmarking cheminformatics tools for automated reaction data extraction.
• Supporting research in scientific document parsing for chemistry literature.

Strengths

• Contains 15,400 images, providing a substantial scale for model training.
• Includes detailed annotations for chemical reaction diagram parsing.
• Split into 15,000 training and 400 test images, facilitating machine learning workflows.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The source and method of PDF extraction are not detailed, which may affect reproducibility.
• Data may reflect publication bias inherent to the scientific literature sources used.

Provenance

Source

Images extracted from scientific literature PDFs.

Collection Method

Extracted from scientific literature PDFs; specific extraction method not detailed.

Time Range

null

Freshness

Last updated 2026-03-11 07:14:30; freshness should be verified.

Geography

null