Note: This project expects the
@zxaoyou/segmentation_WBCdataset in the root directory. It could, however, be used with other datasets ifdataset.pyis replaced with some code to parse the replacement dataset.
Generation of artificial white blood cell micrographs using Pyhon
This is a project I did for school. It adapts the work of Scalbert et al. to generating white blood cells (WBCs). Use python3 <script> --help to view expected parameters for each script. Details can be found in report.Rmd, which can be turned into a PDF using RStudio. A pretrained mask model is provided in models, as well as some example masks and images in images.
Unfortunately, the texture transfer routine used in this project uses up a huge amount of memory, severely limiting the size of the images that can be generated. We may be able to try something like @chuanli11/CNNMRF, where we would operate on channel-wise batches of patches, and potentially convert the normalized cross correlation to a convolution with style patches, followed by normalization. Unlike that project, however, we need to consider the semantic layers, and I don't think that we can just concatenate them with each batch.
A list of references used in this project can be found in report.Rmd.