VISTA-Cell

The VISTA2D is a cell segmentation training and inference pipeline for cell imaging [ Blog ]. A pretrained model was trained on collection of 15K public microscopy images. The data collection and training can be reproduced following the download_preprocessor/ . Alternatively, the model can be retrained on your own dataset. The pretrained vista2d model achieves good performance on diverse set of cell types, microscopy image modalities, and can be further finetuned if necessary. The codebase utilizes several components from other great works including SegmentAnything and Cellpose, which must be pip installed as dependencies. Vista2D codebase follows MONAI bundle format and its specifications.