0
$\begingroup$

I have a dataset with roughly 800 images that are classified in 18 classes.
The classes are spread unevenly, with some classes having 30 images and others having 5.

In order to increase my dataset,I've decided to use image augmentation modifying each image a little,making 20 new images for every image.

I then decided to use my created images as my training set and my original images as my validation set.

Due to unavailability of a GPU,I couldn't train it a lot,but I ended up with around 50% success rate on a training set and 30% on the validation set.

Was the decision of only using my original dataset as validation a good one? If not,why?

$\endgroup$

1 Answer 1

2
$\begingroup$

I don't think it is wise. Your intention to do validation on your real data is correct. But the way you have it now your model will be prevented from training on data that is from the same distribution as what you actually want to predict.

It is best to first split the images into training and validation sets, then do data augmentation on the training set.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.