Flower Simulation of Training Weldon on Camelyon

[yan-gao-GY]

Proposal

Medical imaging is a suitable case using FL over different hospitals. Camelyon dataset contains whole-slide images (WSI) of hematoxylin and eosin (H&E) stained lymph node sections. The data is labelled with “normal” or “tumor”.

Here we want to provide an example of training a Weldon model on this dataset with FedAvg using Flower. Weldon model is a kind of convolutional neural network (CNN) used for weakly supervised learning.

Changelog entry