datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
capstone-figures
Capstone Figure Dataset
Scientific figure images extracted from ICLR 2024, ICLR 2025, NeurIPS 2023/2024,
ICML 2023/2024 papers. Used for multidimensional figure quality assessment.
Files
Each image is named {VENUE}_{YEAR}_{ID:05d}_fig_{N}.png.
Annotation data
JSONL annotation files are available in the companion GitHub repo.
Paper source
Papers can be re-downloaded with download_papers.py from the GitHub repo.
EXSCLAIM-figure-separator
Figure Separation Training Data For EXSCLAIM
Data Fields
One record refers to one figure's information from the paper.
Journal: String. The journal that this article was taken from (i.e., arXiv, Nature)
ID: String. The ID of the article that this figure comes from, as well as it's figure number.
URL: String. The ID to the article.
Image: PIL.Image.Image. A PIL instance of the image.
Info:
Master Image: List[].
Subfigure Label: List[].
Scale Bar Label: List[].… See the full description on the dataset page: https://huggingface.co/datasets/lwashington3/EXSCLAIM-figure-separator.