What are some of the commonly used image processing techniques for multiclass image classification?

Question

I'm working on multiclass skin disease image classification(caused by bacteria and fungus). Some of the sample images are shown below.

Images contain different background as shown in image_1 and image_3. Also some of the images do not have a background as shown in image_2.

I'm using CNN architectures for multi-class image classification. I have two questions.

1. what image processing techniques are necessary (apart from image resizing) so that the overall performance of the model improves?

2. Does having the same background for all the images (for ex: a white or black ground) helps in training the model?

Answer 1

Depending on the model you are using, rescaling the image can help you in bettering the performances. It means passing from [0, 255] range to [0, 1] or [-1,1] range for pixel values. It depends on the network you are using. In Tensorflow (the module that I use to perform DL calculations) you can use this function. You could use also Data Augmentation in your preprocessing layers like reported here. It can help in bettering the performances and avoiding overfitting.