search

timqqt / pytorch-zoom-in-network

Source code for "Efficient Classification of Very Large Images with Tiny Objects"
25 stars 9 forks source link

about data preprocess #3

Open Christian-lyc opened 1 year ago

Christian-lyc commented 1 year ago

Hi, thank you for your code. Just have a question when I go through your code. I didn't find any preprocess, data augmentation or normalization in the code. I'm wondering how you do data preprocessing.

timqqt commented 1 year ago

Since we assume the input image is very large, we do not augment the input image. Instead, we augment the selected patches. I divide the image pixels by 255.0 for normalization.