0

I want to create a deep learning model to classify images. My dataset has around 400 classes and the classes have different number of images..

  • How can I train the deep learning network on unbalanced datasets of images?

I will use data augmentation to increase the amount of data. Also I will apply oversampling..

  • When should I apply oversampling before or after splitting the images into training, testing, validation sets?

  • Should I make oversampling manually?

Ferdi
  • 4,882
  • 7
  • 42
  • 62
N.IT
  • 101
  • 3
  • what's the use of oversampling testing and validation anyways. I'd over sample training, but let test and validation be bit larger than they'd be with no oversampling.. – Minsky Jan 25 '21 at 18:55

1 Answers1

1

Unbalanced classes are almost certainly not a problem, and oversampling will not solve a non-problem: Are unbalanced datasets problematic, and (how) does oversampling (purport to) help?

(See here for a general motivation for this answer. See here for a motivation for short answers. Longer answers are always welcome.)

Stephan Kolassa
  • 95,027
  • 13
  • 197
  • 357