CoNSeP, the full name of which is Colorectal Nuclear Segmentation and Phenotypes, is a rectal cancer tumor marker segmentation and classification dataset, consisting of 41 H&E-stained image slices, each with a size of 1,000×1,000 pixels and a magnification of 40 times. The images were extracted from 16 whole-slide images of colorectal adenocarcinoma (CRA), each belonging to a single patient, and scanned using an Omnyx VL120 scanner at the Department of Pathology, University Hospitals Coventry and Warwickshire, UK. This dataset was released by the University of Warwick in 2019.