def __init__(self,
root,
output_size,
imset='2017/train.txt',
clip_n=3,
max_obj_n=11):
self.root = root
self.clip_n = clip_n
self.output_size = output_size
self.max_obj_n = max_obj_n
dataset_path = os.path.join(root, 'ImageSets', imset)
self.dataset_list = list()
with open(os.path.join(dataset_path), 'r') as lines:
for line in lines:
dataset_name = line.strip()
if len(dataset_name) > 0:
self.dataset_list.append(dataset_name)
self.random_horizontal_flip = mytrans.RandomHorizontalFlip(0.3)
self.color_jitter = TF.ColorJitter(0.1, 0.1, 0.1, 0.02)
self.random_affine = mytrans.RandomAffine(degrees=15,
translate=(0.1, 0.1),
scale=(0.95, 1.05),
shear=10)
self.random_resize_crop = mytrans.RandomResizedCrop(
output_size, (0.8, 1), (0.95, 1.05))
self.to_tensor = TF.ToTensor()
self.to_onehot = mytrans.ToOnehot(max_obj_n, shuffle=True)
def __init__(self,
root,
output_size,
dataset_file='meta.json',
clip_n=3,
max_obj_n=11):
self.root = root
self.clip_n = clip_n
self.output_size = output_size
self.max_obj_n = max_obj_n
dataset_path = os.path.join(root, dataset_file)
with open(dataset_path, 'r') as json_file:
meta_data = json.load(json_file)
self.dataset_list = list(meta_data['videos'])
self.dataset_size = len(self.dataset_list)
self.random_horizontal_flip = mytrans.RandomHorizontalFlip(0.3)
self.color_jitter = TF.ColorJitter(0.1, 0.1, 0.1, 0.02)
self.random_affine = mytrans.RandomAffine(degrees=15,
translate=(0.1, 0.1),
scale=(0.95, 1.05),
shear=10)
self.random_resize_crop = mytrans.RandomResizedCrop(
output_size, (0.3, 0.5), (0.95, 1.05))
self.to_tensor = TF.ToTensor()
self.to_onehot = mytrans.ToOnehot(max_obj_n, shuffle=True)
def trans(is_training = True):
transforms = []
transforms.append(T.ToTensor())
if is_training:
transforms.append(T.RandomHorizontalFlip(0.5))
return T.Compose(transforms)
def make_coco_transforms(image_set):
normalize = T.Compose([T.ToTensor()])
if image_set == 'train':
return T.Compose([
T.RandomHorizontalFlip(0.5),
normalize,
])
if image_set == 'val':
return T.Compose([
normalize,
])
raise ValueError(f'unknown {image_set}')
def __init__(self, root, output_size, dataset_file='./assets/pretrain.txt', clip_n=3, max_obj_n=11):
self.root = root
self.clip_n = clip_n
self.output_size = output_size
self.max_obj_n = max_obj_n
self.img_list = list()
self.mask_list = list()
dataset_list = list()
with open(os.path.join(dataset_file), 'r') as lines:
for line in lines:
dataset_name = line.strip()
img_dir = os.path.join(root, 'JPEGImages', dataset_name)
mask_dir = os.path.join(root, 'Annotations', dataset_name)
img_list = sorted(glob(os.path.join(img_dir, '*.jpg'))) + sorted(glob(os.path.join(img_dir, '*.png')))
mask_list = sorted(glob(os.path.join(mask_dir, '*.png')))
if len(img_list) > 0:
if len(img_list) == len(mask_list):
dataset_list.append(dataset_name)
self.img_list += img_list
self.mask_list += mask_list
print(f'\t{dataset_name}: {len(img_list)} imgs.')
else:
print(f'\tPreTrain dataset {dataset_name} has {len(img_list)} imgs and {len(mask_list)} annots. Not match! Skip.')
else:
print(f'\tPreTrain dataset {dataset_name} doesn\'t exist. Skip.')
print(myutils.gct(), f'{len(self.img_list)} imgs are used for PreTrain. They are from {dataset_list}.')
self.random_horizontal_flip = mytrans.RandomHorizontalFlip(0.3)
self.color_jitter = TF.ColorJitter(0.1, 0.1, 0.1, 0.03)
self.random_affine = mytrans.RandomAffine(degrees=20, translate=(0.1, 0.1), scale=(0.9, 1.1), shear=10)
self.random_resize_crop = mytrans.RandomResizedCrop(output_size, (0.8, 1))
self.to_tensor = TF.ToTensor()
self.to_onehot = mytrans.ToOnehot(max_obj_n, shuffle=True)
def get_transform(train):
transforms = []
transforms.append(T.ToTensor())
if train:
transforms.append(T.RandomHorizontalFlip(0.5))
return T.Compose(transforms)
# trainloader = torch.utils.data.DataLoader(CSDataSet(args.data_dir, './dataset/list/cityscapes/train.lst', max_iters=args.num_steps*args.batch_size, crop_size=(h, w),
# scale=args.random_scale, mirror=args.random_mirror, mean=IMG_MEAN),
# batch_size=args.batch_size, shuffle=True, num_workers=4, pin_memory=True)
# valloader = torch.utils.data.DataLoader(CSDataSet(args.data_dir, './dataset/list/cityscapes/val.lst', crop_size=(1024, 2048), mean=IMG_MEAN, scale=False, mirror=False),
# batch_size=2, shuffle=False, pin_memory=True)
value_scale = 255
mean = [0.485, 0.456, 0.406]
mean = [item * 255 for item in mean]
std = [0.229, 0.224, 0.225]
std = [item * 255 for item in std]
train_transform = my_trans.Compose([
# my_trans.Resize((args.height, args.width)),
# my_trans.RandScale([0.5, 2.0]),
# my_trans.RandomGaussianBlur(),
my_trans.RandomHorizontalFlip(),
# my_trans.Crop([args.height, args.width],crop_type='rand', padding=mean, ignore_label=255),
my_trans.ToTensor(), # without div 255
my_trans.Normalize(mean=mean, std=std)
])
val_transform = my_trans.Compose([
# my_trans.Resize((args.height, args.width)),
my_trans.ToTensor(), # without div 255
my_trans.Normalize(mean=mean, std=std)
])
data_dir = '/data/zzg/CamVid/'
train_dataset = CamVid(data_dir,
mode='train',
p=None,
transform=train_transform)
default='cuda:0',
help='cpu or cuda:0 or cuda:1')
args = parser.parse_args() if string is None else parser.parse_args(string)
return args
if __name__ == '__main__':
args = parse_args()
wandb.init(config=args, project='dlcv_gan_face')
transform = transforms.Compose([
transforms.Resize(args.img_size),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize([0.5] * 3, [0.5] * 3)
])
train_dataset = Face_Dataset('../hw3_data/face/train', transform)
valid_dataset = Face_Dataset('../hw3_data/face/test', transform)
train_dataloader = DataLoader(train_dataset,
batch_size=args.batch,
shuffle=True,
num_workers=args.num_workers)
valid_dataloader = DataLoader(valid_dataset,
batch_size=args.batch,
num_workers=args.num_workers)
train(args, train_dataloader, valid_dataloader)