Segmentation of Nuclei using Transformer based Architecture

In microscopy-based cellular analysis, nuclei segmentation is an arduous and critical task. It is the starting point for any research in pathology. While there are machines that automate the process of preparing batches of cells and adding compounds to them, there exists a major bottleneck in identifying the characteristics of such cells. The prime motivation of this paper, is to identify already existing solutions in segmenting nuclei. There already exist Deep Learning techniques for segmentation tasks, most of the research work focused on this subject uses some variation of Masked-RCNN or U-Net architectures for instance segmentation, which are based on CNNs. In this paper we intend to work with a model which utilizes transformers in its architecture, and compare the results with its CNN counterpart.