Distribution based MIL pooling filters: experiments on a lymph node metastases dataset

This repository is the official implementation of Distribution based MIL pooling filters: experiments on a lymph node metastases dataset.

There are different pooling filters used in multiple instance learning (MIL) models. This study analyzes performances of MIL models with different pooling filters in distinct tasks on the CAMELYON16 lymph node metastases dataset:

• MIL tasks on a digital histopathology image patches dataset
• CAMELYON16 whole slide image classification

For each section, we have a corresponding folder. The detailed steps for reproducing the results in each section are described inside the folders.