Dataset

Data diversity plays a pivotal role in constructing generalist microscopy image segmentation models. In this challenge, we incorporated the diversity of microscopy images from four dimensions: cell origins, staining methods, microscope types, and cell morphologies. First, the origin of cells in microscopy images varies significantly, as they can derive from diverse tissues or exist within cell cultures under various conditions. This introduces considerable variability, as cells within tissues tend to be densely packed and spatially organized, whereas cells in culture are often sparsely distributed and randomly positioned. Second, the choice of staining methods, such as Jenner-Giemsa in brightfield microscopy or the utilization of specific antibodies in fluorescent microscopy, further contributes to the diversity by highlighting different cellular structures or proteins. Third, the use of different microscope types, such as brightfield, fluorescent, phase-contrast (PC), and differential interference contrast (DIC), introduces substantial differences in image characteristics, textures, and associated artifacts. Fourth, cell morphologies exhibit substantial variations across different cell types. While most cells tend to have a round shape, certain cells may display elongated or irregular shapes.

 Download

• Zenodo
  
• Google Drive
  
• Baidu Netdisk (pw: 2022)

Data license: CC BY-NC-ND