This repository is no longer maintained, all migrated to BSDA.

BRSDA

Bayesian Random Semantic Data Augmentation (BRSDA) algorithm implement in Pytorch。

• Bayesian Random Semantic Data Augmentation for Medical Image Analysis

Introduction

Get Started

Requirements

Run

Result

How to use our method

from .brsda_warp import BRSDAWarp

model = Backbone()
brsda_warp = BRSDAWarp(model, num_classes, multi=10, lambda=0.8)

...
# Training
outputs = brsda_warp(x, is_train=True) 
loss = brsda_warp.get_loss(outputs, targets, criterion, is_train=True)

# forward loss
...

Acknowledgment

• Our code for classification is mainly based on MedMNIST
• Our code for augmentation method if maily based on MedAugment
• Our code for segmentation if maily based on
• Thanks for https://github.com/ajbrock/BigGAN-PyTorch providing code for visualization.

Citation