def __init__(self, path, base_size, crop_size, split, overfit):
self.env = lmdb.open(os.path.join(path, split + ".db"),
subdir=False,
readonly=True,
lock=False,
readahead=False,
meminit=False)
with self.env.begin(write=False) as txn:
self.image_paths = pickle.loads(txn.get(b'__keys__'))
self.path = path
self.split = split
self.crop_size = crop_size
self.base_size = base_size
self.overfit = overfit
if crop_size == -1:
self.scalecrop = tr.ScaleWithPadding(base_size=self.base_size)
else:
self.scalecrop = tr.FixScaleCrop(crop_size=self.crop_size)
if overfit:
self.image_paths = self.image_paths[:1]
if len(self.image_paths) == 0:
raise Exception("No images found in dataset directory")
def transform_val(self, sample):
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=self.args.crop_size),
tr.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)),
tr.ToTensor()])
return composed_transforms(sample)
def transform_val(self, sample): # as appearing in tranform_tr, transform_val also holds the same principle
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=self.args.crop_size), # fixscalecrop, we have to calcualte base_size and crop_size based on argparse
tr.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)),
tr.ToTensor()])
return composed_transforms(sample) # return composed_transform
def transform_val(self, sample):
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=self.crop_size),
tr.Darken(self.cfg),
tr.Normalize(mean=self.data_mean, std=self.data_std),
tr.ToTensor()
])
return composed_transforms(sample)
def transform_val(self, sample):
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=self.crop_size),
tr.Normalize(mean=[0.4911], std=[0.1658]),
tr.ToTensor()
])
transformed = composed_transforms(sample)
transformed['image'] = transformed['image'].unsqueeze(0)
return transformed
def transform_tr(self, sample):
composed_transforms = transforms.Compose([ #将各种transformation组合在一起
# tr.RandomHorizontalFlip(),
# tr.RandomScaleCrop(base_size=self.args.base_size, crop_size=self.args.crop_size, fill=255),
# tr.RandomGaussianBlur(),
tr.FixScaleCrop(crop_size=self.args.crop_size),
tr.Normalize(mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225)),
tr.ToTensor()
])
return composed_transforms(sample)
def transform_tr_part1_1(self, sample):
if self.args.use_small:
composed_transforms = transforms.Compose(
[tr.FixScaleCrop(crop_size=self.args.crop_size)])
else:
composed_transforms = transforms.Compose([
tr.RandomHorizontalFlip(),
tr.RandomScaleCrop(base_size=self.args.base_size,
crop_size=self.args.crop_size)
]) # Zhiwei
return composed_transforms(sample)
def transform_val(self, sample):
"""Image transformations for validation"""
targs = self.transform
method = targs["method"]
pars = targs["parameters"]
composed_transforms = transforms.Compose([
tr.FixedResize(size=pars["outSize"]),
tr.FixScaleCrop(cropSize=pars["outSize"]),
tr.Normalize(mean=pars["mean"], std=(pars["std"])),
tr.ToTensor()])
return composed_transforms(sample)
def transform_train(self, sample):
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=self.crop_size),
tr.RandomHorizontalFlip(),
tr.RandomGaussianBlur(),
tr.Normalize(mean=[0.4911], std=[0.1658]),
tr.ToTensor()
])
transformed = composed_transforms(sample)
transformed['image'] = transformed['image'].unsqueeze(0)
return transformed
def transform_val(self, sample):
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=self.args.crop_size),
tr.Normalize(mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225)),
tr.ToTensor()
])
transformed = composed_transforms(sample)
transformed['imgId'] = sample['imgId']
transformed['resolution'] = sample['image'].size
return transformed
def transform_val(self, sample):
"""Transformations for images
sample: {image:img, annotation:ann}
Note: the mean and std is from imagenet
"""
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=self.args.crop_size, fill=0),
tr.Normalize(mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225)),
tr.ToTensor()
])
return composed_transforms(sample)
def transform_tr(self, sample):
#print(sample)
composed_transforms = transforms.Compose([
tr.RandomHorizontalFlip(),
tr.FixScaleCrop(crop_size=self.args.crop_size),
#tr.RandomScaleCrop(base_size=self.args.base_size, crop_size=self.args.crop_size),
tr.RandomGaussianBlur(),
tr.Normalize(mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225)),
tr.ToTensor(),
tr.Lablize(high_confidence=self.args.high_confidence)
])
return composed_transforms(sample)
def transform_val_part1(self, sample):
if self.args.enable_test and self.args.enable_test_full:
return sample
else:
if self.args.enable_adjust_val:
composed_transforms = transforms.Compose([
tr.AutoAdjustSize(factor=self.args.adjust_val_factor,
fill=254)
])
else:
composed_transforms = transforms.Compose(
[tr.FixScaleCrop(crop_size=self.args.crop_size)])
return composed_transforms(sample)
def transform_ts(self, sample):
"""
composed transformers for testing dataset
:param sample: {'image': image, 'label': label}
:return:
"""
composed_transforms = transforms.Compose([
ct.FixScaleCrop(crop_size=self.crop_size),
ct.Normalize(mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225)),
ct.ToTensor()
])
return composed_transforms(sample)
def __init__(self, env, paths, crop_size, include_labels=False):
self.env = env
self.paths = paths
self.crop_size = crop_size
self.include_labels = include_labels
self.base_size = 512
if crop_size == -1:
self.scalecrop = tr.ScaleWithPadding(base_size=self.base_size)
self.scalecrop_image_only = tr.ScaleWithPaddingImageOnly(
base_size=self.base_size)
else:
self.scalecrop = tr.FixScaleCrop(crop_size=self.crop_size)
self.scalecrop_image_only = tr.FixScaleCropImageOnly(
crop_size=self.crop_size)