How to do image data augmentation with k-fold cross validation

[user_xyz]

Hello everyone, I was wondering if there was a solution on doing image data augmentation on just the training data. I have the k-fold setup

for fold,(train_idx,test_idx) in enumerate(kfold.split(dataset)):
  print('------------fold no---------{}----------------------'.format(fold))
  train_subsampler = torch.utils.data.SubsetRandomSampler(train_idx)
  test_subsampler = torch.utils.data.SubsetRandomSampler(test_idx)
 
  trainloader = torch.utils.data.DataLoader(
                      dataset, 
                      batch_size=batch_size, sampler=train_subsampler)
  testloader = torch.utils.data.DataLoader(
                      dataset,
                      batch_size=batch_size, sampler=test_subsampler)
 
  model.apply(reset_weights)
 
  for epoch in range(1, epochs + 1):
    train(fold, model, device, trainloader, optimizer, epoch)
    test(fold,model, device, testloader)

But during those folds, I am unable to find a way to augment just the training data just before the trainloader.

I see that other posts from years ago has not been answered. Has there been any update to this?

How to augment train data during k-Fold cross validation

I am trying to use data augmentation for each of the epoch on train set, but I also need the filenames of testloader for later. So, I used a custom ImageFolderWithPaths to generate tuple for image, label, path. But when combined with a wrapper dataset to build using augmentation have some issues. import os import torch from torch import nn from torch.utils.data import DataLoader, ConcatDataset from torchvision import transforms import torchvision from sklearn.model_selection import KFold from…

Correct data augmentation with k-fold cross validation

Hi! I’m performing 10 k-fold cross validation on my neural network model. I’ve done the splits (k=10: 90% train, 10% validation), but It’s not clear for me if I should apply transforms (random horizontal flip, random rotation, etc) only on training set or both training and validation. What is the correct way to do it? Thanks!

Thank you!

[ptrblck]

This post might be helpful.