search

Xinyu-Xiang / DIVFusion

DIVFusion: Darkness-free infrared and visible image fusion
105 stars 10 forks source link

About data augmentation #4

Open realqikuaiqian opened 1 year ago

realqikuaiqian commented 1 year ago

Hi author, I have read your paper and code, and it's really impressive work. I have a question about the data augmentation process. Could you please explain in detail how you crop the images into multiple image patches for training?

Xinyu-Xiang commented 1 year ago

crop the images by the function of python, i.e. 'y = random.randint(0, h - patch_size_y - 1)', which try to crop them randomly and the size is 224*224

Xinyu-Xiang commented 1 year ago

you can see the code in decomposition.py around line 170

realqikuaiqian commented 1 year ago

I've understood, thanks for your answers.