K-POP: Predicting Distance to Focal Plane for Kato-Katz Prepared Microscopy Slides Using Deep Learning
This repository contains all code produced to write my master's thesis available here.
The model code which is integrated as a git submodule is hosted on HuggingFace.
To fully initialize all submodules run the following after cloning:
git submodule update --init --recursive
The machine learning pipeline works as follows:
- Using the code in
annotation-preprocessing
metadata and images are extracted from a database. The data is prepared to be usable by the tool Focus Annotator. - Data is annotated using Focus Annotator (folder:
focus_annotator
) - Annotations are post processed to csv format to be easier to load with pandas (folder:
data-preprocessing
). - The models are trained and evaluated in
models
.