The images in the Kumar dataset come from The Cancer Genome Atlas (TCGA). Each image is 1000 x 1000 pixels in size and is cropped from a densely populated area of tissue.

To further ensure the richness of the cell nucleus appearance, the dataset covers seven different organs, namely breast, liver, kidney, prostate, bladder, colon and stomach, including benign and diseased tissue samples. The dataset contains 30 images and 22,000 cell nucleus boundary annotations.