This deposition contains an annotated image segmentation datasets and pretrained models. This dataset includes annotated cell images obtained from the Human Protein Atlas (http://www.proteinatlas.org), each image contains 4 channels (Microtubules, ER, Nuclei and Protein of Interest). The cells in each image are annotated with polygons and saved into GeoJSON format produced with Kaibu(https://kaibu.org) annotation tool. hpa_cell_segmentation_dataset_v2_512x512_4train_159test.zip is an example dataset for running a deep learning-based interactive annotation tools in ImJoy (https://github.com/imjoy-team/imjoy-interactive-segmentation). hpa_dataset_v2.zip is a full annotate image segmentation dataset other files named *.pth are pytorch model weights which can be loaded via our deep learning library in Python: https://github.com/CellProfiling/HPA-Cell-Segmentation