def val(opt):
image_1_path = opt.image1_path
image_2_path = opt.image2_path
A_img = Image.open(image_1_path).convert('RGB')
B_img = Image.open(image_2_path).convert('RGB')
trans = transform()
A = trans(A_img).unsqueeze(0)
B = trans(B_img).unsqueeze(0)
# dataset = create_dataset(opt) # create a dataset given opt.dataset_mode and other options
model = create_model(
opt) # create a model given opt.model and other options
model.setup(
opt) # regular setup: load and print networks; create schedulers
save_path = opt.results_dir
mkdir(save_path)
model.eval()
data = {}
data['A'] = A
data['B'] = B
data['A_paths'] = [image_1_path]
model.set_input(data) # unpack data from data loader
pred = model.test(val=False) # run inference return pred
img_path = [image_1_path] # get image paths
save_images(pred, save_path, img_path)
def main():
# step1: opt
opt = TestOptions().parse()
# step2: data
data_loader = CreateDataLoader(opt)
dataset = data_loader.load_data()
# step3: model
model = create_model(opt)
model.setup(opt)
# step4: web ,在test中不使用visdom,而是使用html
web_dir = os.path.join(opt.results_dir, opt.name,
'{}_{}'.format(opt.phase, opt.epoch))
title = 'Experiment = {}, Phase = {}, Epoch = {}'.format(
opt.name, opt.phase, opt.epoch)
webpage = HTML(web_dir, title=title)
# test with eval mode. This only affects layers like batchnorm and dropout.
# pix2pix: we use batchnorm and dropout in the original pix2pix. You can experiment it with and without eval() mode.
# CycleGAN: It should not affect CycleGAN as CycleGAN uses instancenorm without dropout.
model.eval()
for i, data in enumerate(dataset):
if i > opt.num_test:
break
model.set_input(data)
model.test()
visuals = model.get_current_visuals()
img_path = model.get_images_paths()
if i % 5 == 0:
print('processing {:0>4d}-th image...{}'.format(i, img_path))
save_images(webpage,
visuals,
img_path,
aspect_ratio=opt.aspect_ratio,
width=opt.display_winsize)
webpage.save()
def val(opt):
# create a dataset given opt.dataset_mode and other options
dataset = create_dataset(opt)
# create a model given opt.model and other options
model = create_model(opt)
# regular setup: load and print networks; create schedulers
model.setup(opt)
save_path = opt.results_dir
mkdir(save_path)
model.eval()
for i, data in enumerate(dataset):
if i >= opt.num_test: # only apply our model to opt.num_test images.
break
model.set_input(data) # unpack data from data loader
pred = model.test(val=False) # run inference return pred
img_path = model.get_image_paths() # get image paths
if i % 5 == 0: # save images to an HTML file
print('processing (%04d)-th image... %s' % (i, img_path))
save_images(pred, save_path, img_path)
test_loss_iter = []
epoch_iter = 0
conf_mat = np.zeros(
(dataset.dataset.num_labels, dataset.dataset.num_labels),
dtype=np.float)
with torch.no_grad():
for i, data in enumerate(dataset):
model.set_input(data)
model.forward()
model.get_loss()
epoch_iter += opt.batch_size
gt = model.label.cpu().int().numpy()
_, pred = torch.max(model.output.data.cpu(), 1)
pred = pred.float().detach().int().numpy()
save_images(save_dir, model.get_current_visuals(),
model.get_image_names(), model.get_image_oriSize(),
opt.prob_map)
# Resize images to the original size for evaluation
image_size = model.get_image_oriSize()
oriSize = (image_size[0].item(), image_size[1].item())
gt = np.expand_dims(cv2.resize(np.squeeze(gt, axis=0),
oriSize,
interpolation=cv2.INTER_NEAREST),
axis=0)
pred = np.expand_dims(cv2.resize(np.squeeze(pred, axis=0),
oriSize,
interpolation=cv2.INTER_NEAREST),
axis=0)
conf_mat += confusion_matrix(gt, pred, dataset.dataset.num_labels)
test_loss_iter = []
epoch_iter = 0
conf_mat = np.zeros(
(dataset.dataset.num_labels, dataset.dataset.num_labels),
dtype=np.float)
with torch.no_grad():
for i, data in enumerate(dataset):
model.set_input(data)
model.forward()
model.get_loss()
epoch_iter += opt.batch_size
gt = model.label.cpu().int().numpy()
_, pred = torch.max(model.output.data.cpu(), 1)
pred = pred.float().detach().int().numpy()
save_images(save_dir, model.get_current_visuals(),
model.get_image_names(), model.get_image_oriSize())
# Resize images to the original size for evaluation
image_size = model.get_image_oriSize()
oriSize = (image_size[0].item(), image_size[1].item())
gt = np.expand_dims(cv2.resize(np.squeeze(gt, axis=0),
oriSize,
interpolation=cv2.INTER_NEAREST),
axis=0)
pred = np.expand_dims(cv2.resize(np.squeeze(pred, axis=0),
oriSize,
interpolation=cv2.INTER_NEAREST),
axis=0)
conf_mat += confusion_matrix(gt, pred, dataset.dataset.num_labels)
test_loss_iter.append(model.loss_segmentation)
default=512,
help='crop size of test images',
type=int)
parser.add_argument('--name',
default='nature2painting',
help='name of experiment, also where model is stored',
type=str)
args = parser.parse_args()
dataroot = args.dataroot
results_dir = args.result_dir
load_size = args.load_size
crop_size = args.crop_size
name = args.name
dataset = create_dataset(dataroot, batch_size=1, phase='test', load_size=load_size, crop_size=crop_size, serial_batches=True, input_nc=3, output_nc=3, no_flip = True)
model = create_model(isTrain=False, name=name, model='test')
model.setup()
for i, data in enumerate(tqdm(dataset)):
model.set_input(data)
model.test()
visuals = model.get_current_visuals()
img_path = model.get_image_paths()
save_images(visuals, results_dir, img_path, name=str(i))
import pdb
# modify some of the arguments
opt = TestOptions().parse()
opt.nThreads = 1 # test code only supports nThreads = 1
opt.batchSize = 1 # test code only supports batchSize = 1
opt.serial_batches = True # no shuffle
opt.no_flip = True # no flip
data_loader = CreateDataLoader(opt)
dataset = data_loader.load_data(
) # load_date actually returns the dataloader provided by pytorch
model = create_model(opt)
# test
for i, data in enumerate(dataset):
if i >= opt.ntest:
break
print 'Testing progress: {}/{}'.format(i + 1, len(dataset))
model.set_input(data)
model.test(
) # same implementation as model.forward(), but without gradient backprop
visuals = model.get_current_visuals() # returns real_A, fake_B, and real_B
img_path = model.get_image_paths()
save_path = os.path.join(opt.results_dir, opt.name)
save_images(visuals, save_path, img_path)
from models import create_model
from util.util import save_images
if __name__ == '__main__':
opt = TestOptions().parse() # get test options
# hard-code some parameters for test
opt.num_threads = 0 # test code only supports num_threads = 0
opt.batch_size = 1 # test code only supports batch_size = 1
opt.serial_batches = True # disable data shuffling; comment this line if results on randomly chosen images are needed.
opt.no_flip = True # no flip; comment this line if results on flipped images are needed.
dataset = create_dataset(
opt) # create a dataset given opt.dataset_mode and other options
model = create_model(
opt) # create a model given opt.model and other options
model.setup(
opt) # regular setup: load and print networks; create schedulers
# test with eval mode. This only affects layers like batchnorm and dropout.
# For [pix2pix]: we use batchnorm and dropout in the original pix2pix. You can experiment it with and without eval() mode.
img_dir = os.path.join(opt.results_dir, opt.name,
'{}_{}'.format(opt.phase, opt.epoch))
if opt.eval:
model.eval()
for i, data in enumerate(dataset):
if i >= opt.num_test: # only apply our model to opt.num_test images.
break
model.set_input(data) # unpack data from data loader
model.test() # run inference
visuals = model.get_current_visuals() # get image results
img_path = model.get_image_paths() # get image paths
save_images(visuals, img_path, img_dir)
opt.batch_size = 1 # test code only supports batch_size = 1
opt.serial_batches = True # disable data shuffling; comment this line if results on randomly chosen images are needed.
opt.no_flip = True # no flip; comment this line if results on flipped images are needed.
dataset = create_dataset(
opt) # create a dataset given opt.dataset_mode and other options
model = CycleGANModel(
opt) # create a model given opt.model and other options
model.setup(
opt) # regular setup: load and print networks; create schedulers
# create results dir
image_dir = create_results_dir(opt)
# test with eval mode. This only affects layers like batchnorm and dropout.
# For [CycleGAN]: It should not affect CycleGAN as CycleGAN uses instancenorm without dropout.
if opt.eval:
model.eval()
for i, data in enumerate(dataset):
if i >= opt.num_test: # only apply our model to opt.num_test images.
break
model.set_input(data) # unpack data from data loader
model.test() # run inference
visuals = model.get_current_visuals() # get image results
img_path = model.get_image_paths() # get image paths
if i % 5 == 0: # save images to an HTML file
print('processing (%04d)-th image... %s' % (i, img_path))
save_images(opt,
image_dir,
visuals,
img_path,
aspect_ratio=opt.aspect_ratio,
width=opt.display_winsize)