File size: 3,016 Bytes
998bb30 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 |
import argparse
from torch.utils.data import DataLoader
from utils_ import *
parser = argparse.ArgumentParser(description='Clip-based Generative Networks')
parser.add_argument('--test_dir', default='', help='Testing Data')
parser.add_argument('--batch_size', type=int, default=10, help='Batch Size')
parser.add_argument('--model_t',type=str, default= 'all', help ='Model under attack : vgg16, vgg19, ..., dense121')
parser.add_argument('--label_flag', type=str, default='N8', help='Label nums: N8, C20, C50, ...')
parser.add_argument('--finetune', action='store_true', help='Finetune for single class attack')
parser.add_argument('--finetune_class', type=int, help='Class id to be finetuned')
args = parser.parse_args()
print(args)
n_class = 1000
# GPU
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
print(device)
dic = dict()
if args.model_t == 'all':
model_name_list = ['vgg16', 'googlenet', 'incv3', 'res152', 'dense121', 'incv4', 'inc_res_v2', 'adv_incv3', 'ens_inc_res_v2', 'res50_sin', 'res50_sin_in', 'res50_sin_fine_in']
elif args.model_t == 'robust':
model_name_list = ['adv_incv3', 'ens_inc_res_v2', 'res50_sin', 'res50_sin_in', 'res50_sin_fine_in']
elif args.model_t == 'normal':
model_name_list = ['vgg16', 'googlenet', 'incv3', 'res152', 'dense121', 'incv4', 'inc_res_v2']
else:
model_name_list = [args.model_t]
for model_name in model_name_list:
model_t = load_model(model_name)
model_t = model_t.to(device)
model_t.eval()
# Input dimensions: Inception takes 3x299x299
if model_name in ['incv3', 'incv4', 'inc_res_v2', 'adv_incv3', 'ens_inc_res_v2']:
img_size = 299
else:
img_size = 224
# Setup-Data
data_transform = transforms.Compose([
transforms.Resize(img_size),
transforms.ToTensor(),
])
if args.finetune:
class_ids = np.array([args.finetune_class])
else:
class_ids = get_classes(args.label_flag)
# Evaluation
sr = np.zeros(len(class_ids))
for idx in range(len(class_ids)):
test_dir = '{}_t{}'.format(args.test_dir, class_ids[idx])
target_acc = 0.
target_test_size = 0.
test_set = datasets.ImageFolder(test_dir, data_transform)
test_loader = torch.utils.data.DataLoader(test_set, batch_size=args.batch_size, shuffle=False, num_workers=4,
pin_memory=True)
for i, (img, _) in enumerate(test_loader):
img = img.to(device)
adv_out = model_t(normalize(img.clone().detach()))
target_acc += torch.sum(adv_out.argmax(dim=-1) == (class_ids[idx])).item()
target_test_size += img.size(0)
sr[idx] = target_acc / target_test_size
print('sr: {}'.format(sr))
print('model:{} \t target acc:{:.2%}\t target_test_size:{}'.format(model_name, sr.mean(), target_test_size))
dic[model_name] = sr.mean() * 100
print(dic)
|