Prerequisites:
Python3
PyTorch == 0.4.0/0.4.1 (with suitable CUDA and CuDNN version)
torchvision >= 0.2.1
argparse
PIL

Stanford dogs can be found here: http://vision.stanford.edu/aditya86/ImageNetDogs/
MIT Indoors 67 can be found here: http://web.mit.edu/torralba/www/indoor.html
Stanford cars can be found here: http://ai.stanford.edu/~jkrause/cars/car_dataset.html
CUB-200-2011 can be found here: http://www.vision.caltech.edu/visipedia/CUB-200-2011.html
Caltech-256 can be found here: http://www.vision.caltech.edu/Image_Datasets/Caltech256/


run this command to run this code: ($L^2$+BSS)

python BSS.py --trainpath 'trainpath' --testpath 'testpath' --classnum 120

change 'trainpath' to path of your training dataset, 'testpath' to path of your testing dataset, classnum to your num of class in the dataset you choose.