【transformer】【pytorch】DeiT的数据增强
1 main中的相关参数
#函数：def get_args_parser():
parser.add_argument('--input-size', default=224, type=int, help='images input size')
#颜色抖动
parser.add_argument('--color-jitter', type=float, default=0.4, metavar='PCT',
help='Color jitter factor (default: 0.4)')
#rand_augment_transform的参数
parser.add_argument('--aa', type=str, default='rand-m9-mstd0.5-inc1', metavar='NAME',
help='Use AutoAugment policy. "v0" or "original". (default: rand-m9-mstd0.5-inc1)'),
#插值方法
parser.add_argument('--train-interpolation', type=str, default='bicubic',
help='Training interpolation (random, bilinear, bicubic default: "bicubic")')
#repeated
parser.add_argument('--repeated-aug', action='store_true')
parser.add_argument('--no-repeated-aug', action='store_false', dest='repeated_aug')
parser.set_defaults(repeated_aug=True)
#下面的是与随机擦除有关的参数
parser.add_argument('--reprob', type=float, default=0.25, metavar='PCT',
help='Random erase prob (default: 0.25)')
parser.add_argument('--remode', type=str, default='pixel',
help='Random erase mode (default: "pixel")')
parser.add_argument('--recount', type=int, default=1,
help='Random erase count (default: 1)')
parser.add_argument('--resplit', action='store_true', default=False,
help='Do not random erase first (clean) augmentation split')
2 datasets中的build_transform函数
没有直接将库中的create_transform返回，也是为了能够对其进行修改。
def build_transform(is_train, args):
resize_im = args.input_size > 32
#用于训练
if is_train:
# this should always dispatch to transforms_imagenet_train
transform = create_transform(
input_size=args.input_size,
is_training=True,
color_jitter=args.color_jitter,
auto_augment=args.aa,
interpolation=args.train_interpolation,
re_prob=args.reprob,
re_mode=args.remode,
re_count=args.recount,
)
if not resize_im:
# replace RandomResizedCropAndInterpolation with
# RandomCrop
transform.transforms[0] = transforms.RandomCrop(#create_transform返回的是一个列表，可以对列表中的函数进行更改
args.input_size, padding=4)
return transform
#测试
t = []
if resize_im:
size = int((256 / 224) * args.input_size)
t.append(
transforms.Resize(size, interpolation=3), # to maintain same ratio w.r.t. 224 images
)
t.append(transforms.CenterCrop(args.input_size))
t.append(transforms.ToTensor())#最后两个不能忘记
t.append(transforms.Normalize(IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD))
return transforms.Compose(t)#形成的列表放入Compose中
3 create_transform函数
来源：transforms_factory.py(timm库)
函数是一个
#下面带井号的都是传入的参数
def create_transform(
input_size,#
is_training=False,#
use_prefetcher=False,
no_aug=False,
scale=None,
ratio=None,
hflip=0.5,
vflip=0.,
color_jitter=0.4,#
auto_augment=None,#arg.aa=rand-m9-mstd0.5-inc1
interpolation='bilinear',#bicubic
mean=IMAGENET_DEFAULT_MEAN,
std=IMAGENET_DEFAULT_STD,
re_prob=0.,#
re_mode='const',#
re_count=1,#
re_num_splits=0,
crop_pct=None,
tf_preprocessing=False,
separate=False):
...
#没用到的就没有写
transform = transforms_imagenet_train(#使用的是在ImageNet数据集上训练后得到的参数
img_size,
scale=scale,
ratio=ratio,
hflip=hflip,
vflip=vflip,
color_jitter=color_jitter,
auto_augment=auto_augment,
interpolation=interpolation,
use_prefetcher=use_prefetcher,
mean=mean,
std=std,
re_prob=re_prob,
re_mode=re_mode,
re_count=re_count,
re_num_splits=re_num_splits,
separate=separate)
4 transforms_imagenet_train函数
来源：同上
1）transform返回列表：
* 能选择是否返回三个还是合并的一个，显然seperate是这个作用；
* 第一个函数是RandomResizedCropAndInterpolation，记得刚才在函数中对其进行了替换，transform[0]；
2）primary_tfl:
* RandomResizedCropAndInterpolation
* RandomHorizontalFlip（可选）
* RandomVerticalFlip（可选）
3）secondary_tfl：
* rand_augment_transform
* ColorJitter
4）final_tfl:
* ToTensor
* Normalize
* RandomErasing
5）as_params:
* translate_const
* img_mean
def transforms_imagenet_train(
img_size=224,
scale=None,
ratio=None,
hflip=0.5,
vflip=0.,
color_jitter=0.4,
auto_augment=None,
interpolation='random',
use_prefetcher=False,
mean=IMAGENET_DEFAULT_MEAN,
std=IMAGENET_DEFAULT_STD,
re_prob=0.,
re_mode='const',
re_count=1,
re_num_splits=0,
separate=False,
):
"""
If separate==True, the transforms are returned as a tuple of 3 separate transforms
for use in a mixing dataset that passes
* all data through the first (primary) transform, called the 'clean' data
* a portion of the data through the secondary transform
* normalizes and converts the branches above with the third, final transform
"""
scale = tuple(scale or (0.08, 1.0)) # default imagenet scale range
ratio = tuple(ratio or (3./4., 4./3.)) # default imagenet ratio range
primary_tfl = [
RandomResizedCropAndInterpolation(img_size, scale=scale, ratio=ratio, interpolation=interpolation)]
if hflip > 0.:
primary_tfl += [transforms.RandomHorizontalFlip(p=hflip)]
if vflip > 0.:
primary_tfl += [transforms.RandomVerticalFlip(p=vflip)]
secondary_tfl = []
if auto_augment:#rand-m9-mstd0.5-inc1
assert isinstance(auto_augment, str)
if isinstance(img_size, tuple):
img_size_min = min(img_size)
else:
img_size_min = img_size
aa_params = dict(
translate_const=int(img_size_min * 0.45),
img_mean=tuple([min(255, round(255 * x)) for x in mean]),
)
if interpolation and interpolation != 'random':#从这里开始，看使用哪个auto_augment
aa_params['interpolation'] = _pil_interp(interpolation)
if auto_augment.startswith('rand'):#yes
secondary_tfl += [rand_augment_transform(auto_augment, aa_params)]
elif auto_augment.startswith('augmix'):
aa_params['translate_pct'] = 0.3
secondary_tfl += [augment_and_mix_transform(auto_augment, aa_params)]
else:
secondary_tfl += [auto_augment_transform(auto_augment, aa_params)]
elif color_jitter is not None:
# color jitter is enabled when not using AA
if isinstance(color_jitter, (list, tuple)):
# color jitter should be a 3-tuple/list if spec brightness/contrast/saturation
# or 4 if also augmenting hue
assert len(color_jitter) in (3, 4)
else:
# if it's a scalar, duplicate for brightness, contrast, and saturation, no hue
color_jitter = (float(color_jitter),) * 3
secondary_tfl += [transforms.ColorJitter(*color_jitter)]
final_tfl = []
if use_prefetcher:#False
# prefetcher and collate will handle tensor conversion and norm
final_tfl += [ToNumpy()]
else:
final_tfl += [
transforms.ToTensor(),
transforms.Normalize(
mean=torch.tensor(mean),
std=torch.tensor(std))
]
if re_prob > 0.:
final_tfl.append(
RandomErasing(re_prob, mode=re_mode, max_count=re_count, num_splits=re_num_splits, device='cpu'))
if separate:
return transforms.Compose(primary_tfl), transforms.Compose(secondary_tfl), transforms.Compose(final_tfl)
else:
return transforms.Compose(primary_tfl + secondary_tfl + final_tfl)
5 rand_augment_transform函数
调用：secondary_tfl += [rand_augment_transform(auto_augment, aa_params)]
auto_augment:rand-m9-mstd0.5-inc1
aa_params: translate_const, img_mean
def rand_augment_transform(config_str, hparams):
"""
Create a RandAugment transform
:param config_str: String defining configuration of random augmentation. Consists of multiple sections separated by
dashes ('-'). The first section defines the specific variant of rand augment (currently only 'rand'). The remaining
sections, not order sepecific determine
'm' - integer magnitude of rand augment
'n' - integer num layers (number of transform ops selected per image)
'w' - integer probabiliy weight index (index of a set of weights to influence choice of op)
'mstd' - float std deviation of magnitude noise applied
'inc' - integer (bool), use augmentations that increase in severity with magnitude (default: 0) #为1表示使用严重程度随幅度增加的增强
Ex 'rand-m9-n3-mstd0.5' results in RandAugment with magnitude 9, num_layers 3, magnitude_std 0.5
'rand-mstd1-w0' results in magnitude_std 1.0, weights 0, default magnitude of 10 and num_layers 2
:param hparams: Other hparams (kwargs) for the RandAugmentation scheme
:return: A PyTorch compatible Transform
"""
#auto_augment:rand-m9-mstd0.5-inc1
#aa_params: translate_const, img_mean
magnitude = _MAX_LEVEL # default to _MAX_LEVEL for magnitude (currently 10)
num_layers = 2 # default to 2 ops per image
weight_idx = None # default to no probability weights for op choice
transforms = _RAND_TRANSFORMS
config = config_str.split('-') #m9, mstd0.5, inc1
assert config[0] == 'rand'
config = config[1:] # mstd0.5, inc1
for c in config:
cs = re.split(r'(\d.*)', c)
if len(cs) < 2:
continue
key, val = cs[:2]
if key == 'mstd':
# noise param injected via hparams for now
hparams.setdefault('magnitude_std', float(val))
elif key == 'inc':
if bool(val):
transforms = _RAND_INCREASING_TRANSFORMS
elif key == 'm':
magnitude = int(val)
elif key == 'n':
num_layers = int(val)
elif key == 'w':
weight_idx = int(val)
else:
assert False, 'Unknown RandAugment config section'
ra_ops = rand_augment_ops(magnitude=magnitude, hparams=hparams, transforms=transforms) #所有的增强方法
choice_weights = None if weight_idx is None else _select_rand_weights(weight_idx) #如果w没有，那么就是None ,否则就使用_RAND_CHOICE_WEIGHTS_的参数
return RandAugment(ra_ops, num_layers, choice_weights=choice_weights) #随机选取操作，ra_ops是传入的方法str列表，num_layers是增强方法的数目，choice_weight是对应的权重参数
_RAND_INCREASING_TRANSFORMS = [
'AutoContrast',
'Equalize',
'Invert',
'Rotate',
'PosterizeIncreasing',
'SolarizeIncreasing',
'SolarizeAdd',
'ColorIncreasing',
'ContrastIncreasing',
'BrightnessIncreasing',
'SharpnessIncreasing',
'ShearX',
'ShearY',
'TranslateXRel',
'TranslateYRel',
#'Cutout' # NOTE I've implement this as random erasing separately
# These experimental weights are based loosely on the relative improvements mentioned in paper.
# They may not result in increased performance, but could likely be tuned to so.
_RAND_CHOICE_WEIGHTS_0 = {
'Rotate': 0.3,
'ShearX': 0.2,
'ShearY': 0.2,
'TranslateXRel': 0.1,
'TranslateYRel': 0.1,
'Color': .025,
'Sharpness': 0.025,
'AutoContrast': 0.025,
'Solarize': .005,
'SolarizeAdd': .005,
'Contrast': .005,
'Brightness': .005,
'Equalize': .005,
'Posterize': 0,
'Invert': 0,
}
6 RandAugment类
class RandAugment:
def __init__(self, ops, num_layers=2, choice_weights=None):
self.ops = ops
self.num_layers = num_layers
self.choice_weights = choice_weights
def __call__(self, img):
# no replacement when using weighted choice
ops = np.random.choice(
self.ops, self.num_layers, replace=self.choice_weights is None, p=self.choice_weights)
for op in ops:
img = op(img)
return img