def __init__(self, **kwargs):
"""
Initialize preprocessing class for training set
Args:
preprocess (String): Keyword to select different preprocessing types
crop_type (String): Select random or central crop
Return:
None
"""
self.transforms = []
self.transforms1 = []
self.preprocess = kwargs['preprocess']
crop_type = kwargs['crop_type']
self.transforms.append(pt.ResizeClip(**kwargs))
if crop_type == 'Random':
self.transforms.append(pt.RandomCropClip(**kwargs))
else:
self.transforms.append(pt.CenterCropClip(**kwargs))
self.transforms.append(pt.SubtractRGBMean(**kwargs))
self.transforms.append(
pt.RandomFlipClip(direction='h', p=0.5, **kwargs))
self.transforms.append(pt.ToTensorClip(**kwargs))
def __init__(self, **kwargs):
"""
Initialize preprocessing class for training set
Args:
preprocess (String): Keyword to select different preprocessing types
crop_type (String): Select random or central crop
Return:
None
"""
self.transforms = []
self.transforms1 = []
self.preprocess = kwargs['preprocess']
crop_type = kwargs['crop_type']
self.clip_mean = np.load('weights/sport1m_train16_128_mean.npy')[0]
self.clip_mean = np.transpose(self.clip_mean, (1, 2, 3, 0))
self.transforms.append(pt.ResizeClip(**kwargs))
self.transforms.append(
pt.SubtractMeanClip(clip_mean=self.clip_mean, **kwargs))
if crop_type == 'Random':
self.transforms.append(pt.RandomCropClip(**kwargs))
else:
self.transforms.append(pt.CenterCropClip(**kwargs))
self.transforms.append(
pt.RandomFlipClip(direction='h', p=0.5, **kwargs))
self.transforms.append(pt.ToTensorClip(**kwargs))
def __init__(self, **kwargs):
crop_shape = kwargs['crop_shape']
crop_type = kwargs['crop_type']
resize_shape = kwargs['resize_shape']
self.transforms = []
if crop_type == 'Random':
self.transforms.append(pt.RandomCropClip(**kwargs))
elif crop_type == 'Center':
self.transforms.append(pt.CenterCropClip(**kwargs))
self.transforms.append(pt.ResizeClip(**kwargs))
self.transforms.append(pt.SubtractRGBMean(**kwargs))
self.transforms.append(pt.ToTensorClip())
def __init__(self, **kwargs):
"""
Initialize preprocessing class for training set
Args:
preprocess (String): Keyword to select different preprocessing types
crop_type (String): Select random or central crop
Return:
None
"""
self.transforms = []
self.transforms.append(pt.ResizeClip(**kwargs))
self.transforms.append(pt.CenterCropClip(**kwargs))
self.transforms.append(pt.SubtractRGBMean(**kwargs))
self.transforms.append(pt.ToTensorClip(**kwargs))
def __call__(self, x):
return pt.ToTensorClip()(x)