Multi-Scale Attention-based Multiple Instance Learning for Classification of Multi-Gigapixel Histology Images

Model implementation of paper [arXiv] by Made Satria Wibawa, Kwok-Wai Lo, Lawrence Young, Nasir Rajpoot

This is a supplementary code and notebook for training and testing the model in MNIST dataset from Multi-Scale Attention-based Multiple Instance Learning for Classification of Multi-Gigapixel Histology Images paper.

Presented in AI-enabled Medical Image Analysis (AIMIA) workshop at ECCV 2022.

Attention score for true positive prediction.
Number below the digit images is the attention score, each row represent n-th attention layer.

Attention score for true negative prediction.
Number below the digit images is the attention score, each row represent n-th attention layer.