All images and data are licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. Please remember to cite us when using the images or classifications.
By clicking the button below, you can download all 26158 annotations corresponding to 15949 different cells and the 959 images available in this database.
Download ISBI 2026 Challenge Dataset.
This dataset classifies image mini-patches into two categories: Background and Non-background. Non-background patches contain relevant cellular content, while background patches correspond to regions without cells or without diagnostically relevant cellular information.
The images and annotations are divided in 4 different folders for train, validation and two test sets.
This dataset evaluates the quality of image mini-patches using two main categories: Suitable and Unsuitable. Unsuitable patches contain no relevant cells, including regions with red blood cells, artifacts, debris, or indistinguishable cell clusters. Suitable patches contain at least one potentially relevant cell.
The Suitable category is further subdivided according to visual quality:
This dataset contains whole-slide images (WSIs) in TIFF format from 117 scanned cervical smear samples obtained from 107 patients. Each slide is annotated with a global diagnostic label based on expert visual inspection of the entire sample. The dataset includes samples with the following diagnostic categories: NILM, ASC-US, LSIL, ASC-H, HSIL-LSIL, HSIL, and SCC.