imagenet_eval.py

import os
import sys
import torch
import argparse
import datetime
sys.path.append(os.path.realpath('..'))

from utils import loader_imgnet, model_imgnet, evaluate

def main(args):
    print(args)
    DEVICE = torch.device("cuda:0")
    time1 = datetime.datetime.now()
    dir_data = args.data_dir
    dir_uap = args.uaps_save
    batch_size = args.batch_size
    model_dimension = 299 if args.model_name == 'inception_v3' else 256
    center_crop = 299 if args.model_name == 'inception_v3' else 224
    loader = loader_imgnet(dir_data, 50000, batch_size, model_dimension,center_crop)

    model = model_imgnet(args.model_name)

    uap = torch.load(dir_uap)
    _, _, _, _, outputs, labels, y_outputs = evaluate(model, loader, uap = uap,batch_size=batch_size,DEVICE = DEVICE)
    print('true image Accuracy:', sum(y_outputs == labels) / len(labels))
    print('adversarial image Accuracy:', sum(outputs == labels) / len(labels))
    print('fooling rate:', 1-sum(outputs == labels) / len(labels))
    print('fooling ratio:', 1-sum(y_outputs == outputs) / len(labels))
    time2 = datetime.datetime.now()
    print("time consumed: ", time2 - time1)

def parse_arguments(argv):
    parser = argparse.ArgumentParser()
    parser.add_argument('--data_dir', default='/../imagent/val/',
                        help='training set directory')
    parser.add_argument('--uaps_save', default='./uaps_save/spgd/spgd_10000_20epoch_250batch.pth',
                        help='training set directory')
    parser.add_argument('--batch_size', type=int, help='', default=250)
    parser.add_argument('--model_name', default='vgg16', help='loss type')
    return parser.parse_args(argv)


if __name__ == '__main__':
    main(parse_arguments(sys.argv[1:]))