DataTorchio/nuclei-segmentation-dsb2018

Nuclei Segmentation (DSB2018, BBBC038)

The training set of the 2018 Data Science Bowl: 670 microscopy images with 29,430 individually segmented nuclei, as polygon annotations in COCO format. Images span fluorescence and brightfield or histology stains, several magnifications, and samples from humans, mice and flies.

Browse it before you download it. Every image, with every nucleus drawn on it, is viewable in the browser at https://datatorch.io/datatorch/nuclei-segmentation-dsb2018. The annotations here are the official per-nucleus masks converted to polygons (12 of the 29,442 masks were too small to outline and were dropped).

Contents

Images	670 (stage1_train)
Nuclei	29,430, one polygon each
Format	images plus one COCO annotations JSON
License	CC0 1.0 (public domain)
Source	https://bbbc.broadinstitute.org/BBBC038

Citation

We used image set BBBC038v1, available from the Broad Bioimage Benchmark Collection.

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

About this mirror

Mirrored by DataTorch, an annotation platform. A step-by-step tutorial that uses this dataset is at https://datatorch.io/resources/nuclei-segmentation-tutorial, and a description of the dataset is at https://datatorch.io/resources/nuclei-segmentation-dataset.