Nuclei Segmentation (DSB2018)

The nuclei segmentation dataset from the 2018 Data Science Bowl, mirrored from the Broad Bioimage Benchmark Collection (BBBC038v1) with every nucleus converted to a browsable polygon annotation. The original release ships one binary mask PNG per nucleus; here each mask is a polygon you can inspect in the annotator, filter, and export.

Contents

Images	670 (the official stage1_train split)
Annotations	29,430 nucleus polygons
Classes	1 (nucleus)
Dataset	stage1-train
Imaging	Fluorescence (DAPI and Hoechst class stains) and brightfield/histology, multiple organisms, magnifications, and acquisition setups

The deliberate variety is the point of this benchmark: it was built to test whether a single model can find nuclei across imaging experiments it has never seen, not just one lab's microscope.

Using it

Browse any image to see the per-nucleus polygons. To download everything (free account required):

• ZIP: use the Download button on this project.
• Python SDK: pip install datatorch, then:

datatorch login
datatorch pull datatorch/nuclei-segmentation-dsb2018

Annotations export in COCO format through the project's export schema.

Conversion notes

Polygons were generated from the official per-nucleus masks (largest external contour per mask). 12 of 29,442 masks were dropped as degenerate (fewer than 3 contour points, single-pixel speckles). Mask identities and image filenames match the original release, so results remain comparable with published work.

License and citation

CC0 1.0 (public domain). The contributors waived copyright; attribution is still good practice. If you use this data, please cite:

Caicedo, J.C., Goodman, A., Karhohs, K.W. et al. Nucleus segmentation across imaging experiments: the 2018 Data Science Bowl. Nature Methods 16, 1247-1253 (2019).

Source: BBBC038v1, Broad Bioimage Benchmark Collection.

Related

• More microscopy datasets on DataTorch
• All public datasets