Foundation Models Embedding Evaluation

This repository was made for experimenting with the embeddings produced by different histopathology foundation models. They were evaluated over the task of glomerulus lesion classification.

Setup

• Create a Python Virtual Environment and install required packages.

python3 -m venv .
./bin/python3 -m pip install -r requirements.txt

Replicating the results

For more options or more specific help, use the flag --help when running any given script.

Feature extraction

./bin/python3 -m embedding.extraction\
     -i /path/to/dataset/ \
     -o /output/path/ \
     --classes class1 class2 class3 ...

EfficientNet-B0 training

./bin/python3 -m scripts.train_effnetb0 

Generating graphs

./bin/python3 -m scripts.cross-validation_performance_graph /path/to/generated/logs/folder/