Train specific classes of dataset

[Cesar_Sotelo (César Sotelo)]

Hi. I’m new on the deep learning world, so I decided to build a model for pneumonia detection for my country. I’ve selected a well known dataset NIH Chest X-ray. It have 14 different classes but I interested only in two classes: Consolidation and Infiltration. I’ll use Data augmentation and transfer learning for the training. Could you please give me an example how train only these 2 clases? I’m thinking about read de csv label database delete images who not have the finding label.

Thank you in advance

[ptrblck]

@Cesar_Sotelo

I’m thinking about read de csv label database delete images who not have the finding label.

This sounds like a valid approach.
If you are not interested in training the other classes at the moment, I would either remove them or create a separate smaller dataset, which only contains the desired class samples.