iicd/xautodl
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Updates

Browse Source
This commit is contained in:
D-X-Y 2021-04-22 19:12:21 +08:00
parent cd253112ee
commit 275831b375
12 changed files with 1127 additions and 596 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
.github/workflows/basic_test.yml vendored
View File

@ -48,5 +48,13 @@ jobs:
ls ls
python --version python --version
python -m pytest ./tests/test_basic_space.py -s python -m pytest ./tests/test_basic_space.py -s
shell: bash
- name: Test Synthetic Data
run: |
python -m pip install pytest numpy
python -m pip install parameterized
python -m pip install torch
python --version
python -m pytest ./tests/test_synthetic.py -s python -m pytest ./tests/test_synthetic.py -s
shell: bash shell: bash

2
.latent-data/NATS-Bench

@ -1 +1 @@
Subproject commit 3a8794322f0b990499a44db1b2cb05ef2bb33851 Subproject commit 33bfb2eb1388f0273d4cc492091b1f983340879b

185
lib/datasets/DownsampledImageNet.py
View File

@ -5,118 +5,133 @@ import os, sys, hashlib, torch
import numpy as np import numpy as np
from PIL import Image from PIL import Image
import torch.utils.data as data import torch.utils.data as data
if sys.version_info[0] == 2: if sys.version_info[0] == 2:
import cPickle as pickle import cPickle as pickle
else: else:
import pickle import pickle
def calculate_md5(fpath, chunk_size=1024 * 1024): def calculate_md5(fpath, chunk_size=1024 * 1024):
md5 = hashlib.md5() md5 = hashlib.md5()
with open(fpath, 'rb') as f: with open(fpath, "rb") as f:
for chunk in iter(lambda: f.read(chunk_size), b''): for chunk in iter(lambda: f.read(chunk_size), b""):
md5.update(chunk) md5.update(chunk)
return md5.hexdigest() return md5.hexdigest()
def check_md5(fpath, md5, **kwargs): def check_md5(fpath, md5, **kwargs):
return md5 == calculate_md5(fpath, **kwargs) return md5 == calculate_md5(fpath, **kwargs)
def check_integrity(fpath, md5=None): def check_integrity(fpath, md5=None):
if not os.path.isfile(fpath): return False if not os.path.isfile(fpath):
if md5 is None: return True return False
else : return check_md5(fpath, md5) if md5 is None:
return True
else:
return check_md5(fpath, md5)
class ImageNet16(data.Dataset): class ImageNet16(data.Dataset):
# http://image-net.org/download-images # http://image-net.org/download-images
# A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasets # A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasets
# https://arxiv.org/pdf/1707.08819.pdf # https://arxiv.org/pdf/1707.08819.pdf
train_list = [ train_list = [
['train_data_batch_1', '27846dcaa50de8e21a7d1a35f30f0e91'], ["train_data_batch_1", "27846dcaa50de8e21a7d1a35f30f0e91"],
['train_data_batch_2', 'c7254a054e0e795c69120a5727050e3f'], ["train_data_batch_2", "c7254a054e0e795c69120a5727050e3f"],
['train_data_batch_3', '4333d3df2e5ffb114b05d2ffc19b1e87'], ["train_data_batch_3", "4333d3df2e5ffb114b05d2ffc19b1e87"],
['train_data_batch_4', '1620cdf193304f4a92677b695d70d10f'], ["train_data_batch_4", "1620cdf193304f4a92677b695d70d10f"],
['train_data_batch_5', '348b3c2fdbb3940c4e9e834affd3b18d'], ["train_data_batch_5", "348b3c2fdbb3940c4e9e834affd3b18d"],
['train_data_batch_6', '6e765307c242a1b3d7d5ef9139b48945'], ["train_data_batch_6", "6e765307c242a1b3d7d5ef9139b48945"],
['train_data_batch_7', '564926d8cbf8fc4818ba23d2faac7564'], ["train_data_batch_7", "564926d8cbf8fc4818ba23d2faac7564"],
['train_data_batch_8', 'f4755871f718ccb653440b9dd0ebac66'], ["train_data_batch_8", "f4755871f718ccb653440b9dd0ebac66"],
['train_data_batch_9', 'bb6dd660c38c58552125b1a92f86b5d4'], ["train_data_batch_9", "bb6dd660c38c58552125b1a92f86b5d4"],
['train_data_batch_10','8f03f34ac4b42271a294f91bf480f29b'], ["train_data_batch_10", "8f03f34ac4b42271a294f91bf480f29b"],
] ]
valid_list = [ valid_list = [
['val_data', '3410e3017fdaefba8d5073aaa65e4bd6'], ["val_data", "3410e3017fdaefba8d5073aaa65e4bd6"],
] ]
def __init__(self, root, train, transform, use_num_of_class_only=None): def __init__(self, root, train, transform, use_num_of_class_only=None):
self.root = root self.root = root
self.transform = transform self.transform = transform
self.train = train # training set or valid set self.train = train # training set or valid set
if not self._check_integrity(): raise RuntimeError('Dataset not found or corrupted.') if not self._check_integrity():
raise RuntimeError("Dataset not found or corrupted.")
if self.train: downloaded_list = self.train_list if self.train:
else : downloaded_list = self.valid_list downloaded_list = self.train_list
self.data = []
self.targets = []
# now load the picked numpy arrays
for i, (file_name, checksum) in enumerate(downloaded_list):
file_path = os.path.join(self.root, file_name)
#print ('Load {:}/{:02d}-th : {:}'.format(i, len(downloaded_list), file_path))
with open(file_path, 'rb') as f:
if sys.version_info[0] == 2:
entry = pickle.load(f)
else: else:
entry = pickle.load(f, encoding='latin1') downloaded_list = self.valid_list
self.data.append(entry['data']) self.data = []
self.targets.extend(entry['labels']) self.targets = []
self.data = np.vstack(self.data).reshape(-1, 3, 16, 16)
self.data = self.data.transpose((0, 2, 3, 1)) # convert to HWC
if use_num_of_class_only is not None:
assert isinstance(use_num_of_class_only, int) and use_num_of_class_only > 0 and use_num_of_class_only < 1000, 'invalid use_num_of_class_only : {:}'.format(use_num_of_class_only)
new_data, new_targets = [], []
for I, L in zip(self.data, self.targets):
if 1 <= L <= use_num_of_class_only:
new_data.append( I )
new_targets.append( L )
self.data = new_data
self.targets = new_targets
# self.mean.append(entry['mean'])
#self.mean = np.vstack(self.mean).reshape(-1, 3, 16, 16)
#self.mean = np.mean(np.mean(np.mean(self.mean, axis=0), axis=1), axis=1)
#print ('Mean : {:}'.format(self.mean))
#temp = self.data - np.reshape(self.mean, (1, 1, 1, 3))
#std_data = np.std(temp, axis=0)
#std_data = np.mean(np.mean(std_data, axis=0), axis=0)
#print ('Std : {:}'.format(std_data))
def __repr__(self): # now load the picked numpy arrays
return ('{name}({num} images, {classes} classes)'.format(name=self.__class__.__name__, num=len(self.data), classes=len(set(self.targets)))) for i, (file_name, checksum) in enumerate(downloaded_list):
file_path = os.path.join(self.root, file_name)
# print ('Load {:}/{:02d}-th : {:}'.format(i, len(downloaded_list), file_path))
with open(file_path, "rb") as f:
if sys.version_info[0] == 2:
entry = pickle.load(f)
else:
entry = pickle.load(f, encoding="latin1")
self.data.append(entry["data"])
self.targets.extend(entry["labels"])
self.data = np.vstack(self.data).reshape(-1, 3, 16, 16)
self.data = self.data.transpose((0, 2, 3, 1)) # convert to HWC
if use_num_of_class_only is not None:
assert (
isinstance(use_num_of_class_only, int)
and use_num_of_class_only > 0
and use_num_of_class_only < 1000
), "invalid use_num_of_class_only : {:}".format(use_num_of_class_only)
new_data, new_targets = [], []
for I, L in zip(self.data, self.targets):
if 1 <= L <= use_num_of_class_only:
new_data.append(I)
new_targets.append(L)
self.data = new_data
self.targets = new_targets
# self.mean.append(entry['mean'])
# self.mean = np.vstack(self.mean).reshape(-1, 3, 16, 16)
# self.mean = np.mean(np.mean(np.mean(self.mean, axis=0), axis=1), axis=1)
# print ('Mean : {:}'.format(self.mean))
# temp = self.data - np.reshape(self.mean, (1, 1, 1, 3))
# std_data = np.std(temp, axis=0)
# std_data = np.mean(np.mean(std_data, axis=0), axis=0)
# print ('Std : {:}'.format(std_data))
def __repr__(self):
return "{name}({num} images, {classes} classes)".format(
name=self.__class__.__name__,
num=len(self.data),
classes=len(set(self.targets)),
)
def __getitem__(self, index): def __getitem__(self, index):
img, target = self.data[index], self.targets[index] - 1 img, target = self.data[index], self.targets[index] - 1
img = Image.fromarray(img) img = Image.fromarray(img)
if self.transform is not None: if self.transform is not None:
img = self.transform(img) img = self.transform(img)
return img, target return img, target
def __len__(self): def __len__(self):
return len(self.data) return len(self.data)
def _check_integrity(self):
root = self.root
for fentry in self.train_list + self.valid_list:
filename, md5 = fentry[0], fentry[1]
fpath = os.path.join(root, filename)
if not check_integrity(fpath, md5):
return False
return True
def _check_integrity(self):
root = self.root
for fentry in (self.train_list + self.valid_list):
filename, md5 = fentry[0], fentry[1]
fpath = os.path.join(root, filename)
if not check_integrity(fpath, md5):
return False
return True
""" """
if __name__ == '__main__': if __name__ == '__main__':

440
lib/datasets/LandmarkDataset.py
View File

@ -20,172 +20,282 @@ import torch.utils.data as data
class LandmarkDataset(data.Dataset): class LandmarkDataset(data.Dataset):
def __init__(
self,
transform,
sigma,
downsample,
heatmap_type,
shape,
use_gray,
mean_file,
data_indicator,
cache_images=None,
):
def __init__(self, transform, sigma, downsample, heatmap_type, shape, use_gray, mean_file, data_indicator, cache_images=None): self.transform = transform
self.sigma = sigma
self.transform = transform self.downsample = downsample
self.sigma = sigma self.heatmap_type = heatmap_type
self.downsample = downsample self.dataset_name = data_indicator
self.heatmap_type = heatmap_type self.shape = shape # [H,W]
self.dataset_name = data_indicator self.use_gray = use_gray
self.shape = shape # [H,W] assert transform is not None, "transform : {:}".format(transform)
self.use_gray = use_gray self.mean_file = mean_file
assert transform is not None, 'transform : {:}'.format(transform) if mean_file is None:
self.mean_file = mean_file self.mean_data = None
if mean_file is None: warnings.warn("LandmarkDataset initialized with mean_data = None")
self.mean_data = None
warnings.warn('LandmarkDataset initialized with mean_data = None')
else:
assert osp.isfile(mean_file), '{:} is not a file.'.format(mean_file)
self.mean_data = torch.load(mean_file)
self.reset()
self.cutout = None
self.cache_images = cache_images
print ('The general dataset initialization done : {:}'.format(self))
warnings.simplefilter( 'once' )
def __repr__(self):
return ('{name}(point-num={NUM_PTS}, shape={shape}, sigma={sigma}, heatmap_type={heatmap_type}, length={length}, cutout={cutout}, dataset={dataset_name}, mean={mean_file})'.format(name=self.__class__.__name__, **self.__dict__))
def set_cutout(self, length):
if length is not None and length >= 1:
self.cutout = CutOut( int(length) )
else: self.cutout = None
def reset(self, num_pts=-1, boxid='default', only_pts=False):
self.NUM_PTS = num_pts
if only_pts: return
self.length = 0
self.datas = []
self.labels = []
self.NormDistances = []
self.BOXID = boxid
if self.mean_data is None:
self.mean_face = None
else:
self.mean_face = torch.Tensor(self.mean_data[boxid].copy().T)
assert (self.mean_face >= -1).all() and (self.mean_face <= 1).all(), 'mean-{:}-face : {:}'.format(boxid, self.mean_face)
#assert self.dataset_name is not None, 'The dataset name is None'
def __len__(self):
assert len(self.datas) == self.length, 'The length is not correct : {}'.format(self.length)
return self.length
def append(self, data, label, distance):
assert osp.isfile(data), 'The image path is not a file : {:}'.format(data)
self.datas.append( data ) ; self.labels.append( label )
self.NormDistances.append( distance )
self.length = self.length + 1
def load_list(self, file_lists, num_pts, boxindicator, normalizeL, reset):
if reset: self.reset(num_pts, boxindicator)
else : assert self.NUM_PTS == num_pts and self.BOXID == boxindicator, 'The number of point is inconsistance : {:} vs {:}'.format(self.NUM_PTS, num_pts)
if isinstance(file_lists, str): file_lists = [file_lists]
samples = []
for idx, file_path in enumerate(file_lists):
print (':::: load list {:}/{:} : {:}'.format(idx, len(file_lists), file_path))
xdata = torch.load(file_path)
if isinstance(xdata, list) : data = xdata # image or video dataset list
elif isinstance(xdata, dict): data = xdata['datas'] # multi-view dataset list
else: raise ValueError('Invalid Type Error : {:}'.format( type(xdata) ))
samples = samples + data
# samples is a dict, where the key is the image-path and the value is the annotation
# each annotation is a dict, contains 'points' (3,num_pts), and various box
print ('GeneralDataset-V2 : {:} samples'.format(len(samples)))
#for index, annotation in enumerate(samples):
for index in tqdm( range( len(samples) ) ):
annotation = samples[index]
image_path = annotation['current_frame']
points, box = annotation['points'], annotation['box-{:}'.format(boxindicator)]
label = PointMeta2V(self.NUM_PTS, points, box, image_path, self.dataset_name)
if normalizeL is None: normDistance = None
else : normDistance = annotation['normalizeL-{:}'.format(normalizeL)]
self.append(image_path, label, normDistance)
assert len(self.datas) == self.length, 'The length and the data is not right {} vs {}'.format(self.length, len(self.datas))
assert len(self.labels) == self.length, 'The length and the labels is not right {} vs {}'.format(self.length, len(self.labels))
assert len(self.NormDistances) == self.length, 'The length and the NormDistances is not right {} vs {}'.format(self.length, len(self.NormDistance))
print ('Load data done for LandmarkDataset, which has {:} images.'.format(self.length))
def __getitem__(self, index):
assert index >= 0 and index < self.length, 'Invalid index : {:}'.format(index)
if self.cache_images is not None and self.datas[index] in self.cache_images:
image = self.cache_images[ self.datas[index] ].clone()
else:
image = pil_loader(self.datas[index], self.use_gray)
target = self.labels[index].copy()
return self._process_(image, target, index)
def _process_(self, image, target, index):
# transform the image and points
image, target, theta = self.transform(image, target)
(C, H, W), (height, width) = image.size(), self.shape
# obtain the visiable indicator vector
if target.is_none(): nopoints = True
else : nopoints = False
if index == -1: __path = None
else : __path = self.datas[index]
if isinstance(theta, list) or isinstance(theta, tuple):
affineImage, heatmaps, mask, norm_trans_points, THETA, transpose_theta = [], [], [], [], [], []
for _theta in theta:
_affineImage, _heatmaps, _mask, _norm_trans_points, _theta, _transpose_theta \
= self.__process_affine(image, target, _theta, nopoints, 'P[{:}]@{:}'.format(index, __path))
affineImage.append(_affineImage)
heatmaps.append(_heatmaps)
mask.append(_mask)
norm_trans_points.append(_norm_trans_points)
THETA.append(_theta)
transpose_theta.append(_transpose_theta)
affineImage, heatmaps, mask, norm_trans_points, THETA, transpose_theta = \
torch.stack(affineImage), torch.stack(heatmaps), torch.stack(mask), torch.stack(norm_trans_points), torch.stack(THETA), torch.stack(transpose_theta)
else:
affineImage, heatmaps, mask, norm_trans_points, THETA, transpose_theta = self.__process_affine(image, target, theta, nopoints, 'S[{:}]@{:}'.format(index, __path))
torch_index = torch.IntTensor([index])
torch_nopoints = torch.ByteTensor( [ nopoints ] )
torch_shape = torch.IntTensor([H,W])
return affineImage, heatmaps, mask, norm_trans_points, THETA, transpose_theta, torch_index, torch_nopoints, torch_shape
def __process_affine(self, image, target, theta, nopoints, aux_info=None):
image, target, theta = image.clone(), target.copy(), theta.clone()
(C, H, W), (height, width) = image.size(), self.shape
if nopoints: # do not have label
norm_trans_points = torch.zeros((3, self.NUM_PTS))
heatmaps = torch.zeros((self.NUM_PTS+1, height//self.downsample, width//self.downsample))
mask = torch.ones((self.NUM_PTS+1, 1, 1), dtype=torch.uint8)
transpose_theta = identity2affine(False)
else:
norm_trans_points = apply_affine2point(target.get_points(), theta, (H,W))
norm_trans_points = apply_boundary(norm_trans_points)
real_trans_points = norm_trans_points.clone()
real_trans_points[:2, :] = denormalize_points(self.shape, real_trans_points[:2,:])
heatmaps, mask = generate_label_map(real_trans_points.numpy(), height//self.downsample, width//self.downsample, self.sigma, self.downsample, nopoints, self.heatmap_type) # H*W*C
heatmaps = torch.from_numpy(heatmaps.transpose((2, 0, 1))).type(torch.FloatTensor)
mask = torch.from_numpy(mask.transpose((2, 0, 1))).type(torch.ByteTensor)
if self.mean_face is None:
#warnings.warn('In LandmarkDataset use identity2affine for transpose_theta because self.mean_face is None.')
transpose_theta = identity2affine(False)
else:
if torch.sum(norm_trans_points[2,:] == 1) < 3:
warnings.warn('In LandmarkDataset after transformation, no visiable point, using identity instead. Aux: {:}'.format(aux_info))
transpose_theta = identity2affine(False)
else: else:
transpose_theta = solve2theta(norm_trans_points, self.mean_face.clone()) assert osp.isfile(mean_file), "{:} is not a file.".format(mean_file)
self.mean_data = torch.load(mean_file)
self.reset()
self.cutout = None
self.cache_images = cache_images
print("The general dataset initialization done : {:}".format(self))
warnings.simplefilter("once")
affineImage = affine2image(image, theta, self.shape) def __repr__(self):
if self.cutout is not None: affineImage = self.cutout( affineImage ) return "{name}(point-num={NUM_PTS}, shape={shape}, sigma={sigma}, heatmap_type={heatmap_type}, length={length}, cutout={cutout}, dataset={dataset_name}, mean={mean_file})".format(
name=self.__class__.__name__, **self.__dict__
)
return affineImage, heatmaps, mask, norm_trans_points, theta, transpose_theta def set_cutout(self, length):
if length is not None and length >= 1:
self.cutout = CutOut(int(length))
else:
self.cutout = None
def reset(self, num_pts=-1, boxid="default", only_pts=False):
self.NUM_PTS = num_pts
if only_pts:
return
self.length = 0
self.datas = []
self.labels = []
self.NormDistances = []
self.BOXID = boxid
if self.mean_data is None:
self.mean_face = None
else:
self.mean_face = torch.Tensor(self.mean_data[boxid].copy().T)
assert (self.mean_face >= -1).all() and (
self.mean_face <= 1
).all(), "mean-{:}-face : {:}".format(boxid, self.mean_face)
# assert self.dataset_name is not None, 'The dataset name is None'
def __len__(self):
assert len(self.datas) == self.length, "The length is not correct : {}".format(
self.length
)
return self.length
def append(self, data, label, distance):
assert osp.isfile(data), "The image path is not a file : {:}".format(data)
self.datas.append(data)
self.labels.append(label)
self.NormDistances.append(distance)
self.length = self.length + 1
def load_list(self, file_lists, num_pts, boxindicator, normalizeL, reset):
if reset:
self.reset(num_pts, boxindicator)
else:
assert (
self.NUM_PTS == num_pts and self.BOXID == boxindicator
), "The number of point is inconsistance : {:} vs {:}".format(
self.NUM_PTS, num_pts
)
if isinstance(file_lists, str):
file_lists = [file_lists]
samples = []
for idx, file_path in enumerate(file_lists):
print(
":::: load list {:}/{:} : {:}".format(idx, len(file_lists), file_path)
)
xdata = torch.load(file_path)
if isinstance(xdata, list):
data = xdata # image or video dataset list
elif isinstance(xdata, dict):
data = xdata["datas"] # multi-view dataset list
else:
raise ValueError("Invalid Type Error : {:}".format(type(xdata)))
samples = samples + data
# samples is a dict, where the key is the image-path and the value is the annotation
# each annotation is a dict, contains 'points' (3,num_pts), and various box
print("GeneralDataset-V2 : {:} samples".format(len(samples)))
# for index, annotation in enumerate(samples):
for index in tqdm(range(len(samples))):
annotation = samples[index]
image_path = annotation["current_frame"]
points, box = (
annotation["points"],
annotation["box-{:}".format(boxindicator)],
)
label = PointMeta2V(
self.NUM_PTS, points, box, image_path, self.dataset_name
)
if normalizeL is None:
normDistance = None
else:
normDistance = annotation["normalizeL-{:}".format(normalizeL)]
self.append(image_path, label, normDistance)
assert (
len(self.datas) == self.length
), "The length and the data is not right {} vs {}".format(
self.length, len(self.datas)
)
assert (
len(self.labels) == self.length
), "The length and the labels is not right {} vs {}".format(
self.length, len(self.labels)
)
assert (
len(self.NormDistances) == self.length
), "The length and the NormDistances is not right {} vs {}".format(
self.length, len(self.NormDistance)
)
print(
"Load data done for LandmarkDataset, which has {:} images.".format(
self.length
)
)
def __getitem__(self, index):
assert index >= 0 and index < self.length, "Invalid index : {:}".format(index)
if self.cache_images is not None and self.datas[index] in self.cache_images:
image = self.cache_images[self.datas[index]].clone()
else:
image = pil_loader(self.datas[index], self.use_gray)
target = self.labels[index].copy()
return self._process_(image, target, index)
def _process_(self, image, target, index):
# transform the image and points
image, target, theta = self.transform(image, target)
(C, H, W), (height, width) = image.size(), self.shape
# obtain the visiable indicator vector
if target.is_none():
nopoints = True
else:
nopoints = False
if index == -1:
__path = None
else:
__path = self.datas[index]
if isinstance(theta, list) or isinstance(theta, tuple):
affineImage, heatmaps, mask, norm_trans_points, THETA, transpose_theta = (
[],
[],
[],
[],
[],
[],
)
for _theta in theta:
(
_affineImage,
_heatmaps,
_mask,
_norm_trans_points,
_theta,
_transpose_theta,
) = self.__process_affine(
image, target, _theta, nopoints, "P[{:}]@{:}".format(index, __path)
)
affineImage.append(_affineImage)
heatmaps.append(_heatmaps)
mask.append(_mask)
norm_trans_points.append(_norm_trans_points)
THETA.append(_theta)
transpose_theta.append(_transpose_theta)
affineImage, heatmaps, mask, norm_trans_points, THETA, transpose_theta = (
torch.stack(affineImage),
torch.stack(heatmaps),
torch.stack(mask),
torch.stack(norm_trans_points),
torch.stack(THETA),
torch.stack(transpose_theta),
)
else:
(
affineImage,
heatmaps,
mask,
norm_trans_points,
THETA,
transpose_theta,
) = self.__process_affine(
image, target, theta, nopoints, "S[{:}]@{:}".format(index, __path)
)
torch_index = torch.IntTensor([index])
torch_nopoints = torch.ByteTensor([nopoints])
torch_shape = torch.IntTensor([H, W])
return (
affineImage,
heatmaps,
mask,
norm_trans_points,
THETA,
transpose_theta,
torch_index,
torch_nopoints,
torch_shape,
)
def __process_affine(self, image, target, theta, nopoints, aux_info=None):
image, target, theta = image.clone(), target.copy(), theta.clone()
(C, H, W), (height, width) = image.size(), self.shape
if nopoints: # do not have label
norm_trans_points = torch.zeros((3, self.NUM_PTS))
heatmaps = torch.zeros(
(self.NUM_PTS + 1, height // self.downsample, width // self.downsample)
)
mask = torch.ones((self.NUM_PTS + 1, 1, 1), dtype=torch.uint8)
transpose_theta = identity2affine(False)
else:
norm_trans_points = apply_affine2point(target.get_points(), theta, (H, W))
norm_trans_points = apply_boundary(norm_trans_points)
real_trans_points = norm_trans_points.clone()
real_trans_points[:2, :] = denormalize_points(
self.shape, real_trans_points[:2, :]
)
heatmaps, mask = generate_label_map(
real_trans_points.numpy(),
height // self.downsample,
width // self.downsample,
self.sigma,
self.downsample,
nopoints,
self.heatmap_type,
) # H*W*C
heatmaps = torch.from_numpy(heatmaps.transpose((2, 0, 1))).type(
torch.FloatTensor
)
mask = torch.from_numpy(mask.transpose((2, 0, 1))).type(torch.ByteTensor)
if self.mean_face is None:
# warnings.warn('In LandmarkDataset use identity2affine for transpose_theta because self.mean_face is None.')
transpose_theta = identity2affine(False)
else:
if torch.sum(norm_trans_points[2, :] == 1) < 3:
warnings.warn(
"In LandmarkDataset after transformation, no visiable point, using identity instead. Aux: {:}".format(
aux_info
)
)
transpose_theta = identity2affine(False)
else:
transpose_theta = solve2theta(
norm_trans_points, self.mean_face.clone()
)
affineImage = affine2image(image, theta, self.shape)
if self.cutout is not None:
affineImage = self.cutout(affineImage)
return affineImage, heatmaps, mask, norm_trans_points, theta, transpose_theta

78
lib/datasets/SearchDatasetWrap.py
View File

@ -6,41 +6,49 @@ import torch.utils.data as data
class SearchDataset(data.Dataset): class SearchDataset(data.Dataset):
def __init__(self, name, data, train_split, valid_split, check=True):
self.datasetname = name
if isinstance(data, (list, tuple)): # new type of SearchDataset
assert len(data) == 2, "invalid length: {:}".format(len(data))
self.train_data = data[0]
self.valid_data = data[1]
self.train_split = train_split.copy()
self.valid_split = valid_split.copy()
self.mode_str = "V2" # new mode
else:
self.mode_str = "V1" # old mode
self.data = data
self.train_split = train_split.copy()
self.valid_split = valid_split.copy()
if check:
intersection = set(train_split).intersection(set(valid_split))
assert (
len(intersection) == 0
), "the splitted train and validation sets should have no intersection"
self.length = len(self.train_split)
def __init__(self, name, data, train_split, valid_split, check=True): def __repr__(self):
self.datasetname = name return "{name}(name={datasetname}, train={tr_L}, valid={val_L}, version={ver})".format(
if isinstance(data, (list, tuple)): # new type of SearchDataset name=self.__class__.__name__,
assert len(data) == 2, 'invalid length: {:}'.format( len(data) ) datasetname=self.datasetname,
self.train_data = data[0] tr_L=len(self.train_split),
self.valid_data = data[1] val_L=len(self.valid_split),
self.train_split = train_split.copy() ver=self.mode_str,
self.valid_split = valid_split.copy() )
self.mode_str = 'V2' # new mode
else:
self.mode_str = 'V1' # old mode
self.data = data
self.train_split = train_split.copy()
self.valid_split = valid_split.copy()
if check:
intersection = set(train_split).intersection(set(valid_split))
assert len(intersection) == 0, 'the splitted train and validation sets should have no intersection'
self.length = len(self.train_split)
def __repr__(self): def __len__(self):
return ('{name}(name={datasetname}, train={tr_L}, valid={val_L}, version={ver})'.format(name=self.__class__.__name__, datasetname=self.datasetname, tr_L=len(self.train_split), val_L=len(self.valid_split), ver=self.mode_str)) return self.length
def __len__(self): def __getitem__(self, index):
return self.length assert index >= 0 and index < self.length, "invalid index = {:}".format(index)
train_index = self.train_split[index]
def __getitem__(self, index): valid_index = random.choice(self.valid_split)
assert index >= 0 and index < self.length, 'invalid index = {:}'.format(index) if self.mode_str == "V1":
train_index = self.train_split[index] train_image, train_label = self.data[train_index]
valid_index = random.choice( self.valid_split ) valid_image, valid_label = self.data[valid_index]
if self.mode_str == 'V1': elif self.mode_str == "V2":
train_image, train_label = self.data[train_index] train_image, train_label = self.train_data[train_index]
valid_image, valid_label = self.data[valid_index] valid_image, valid_label = self.valid_data[valid_index]
elif self.mode_str == 'V2': else:
train_image, train_label = self.train_data[train_index] raise ValueError("invalid mode : {:}".format(self.mode_str))
valid_image, valid_label = self.valid_data[valid_index] return train_image, train_label, valid_image, valid_label
else: raise ValueError('invalid mode : {:}'.format(self.mode_str))
return train_image, train_label, valid_image, valid_label

1
lib/datasets/__init__.py
View File

@ -4,4 +4,5 @@
from .get_dataset_with_transform import get_datasets, get_nas_search_loaders from .get_dataset_with_transform import get_datasets, get_nas_search_loaders
from .SearchDatasetWrap import SearchDataset from .SearchDatasetWrap import SearchDataset
from .synthetic_adaptive_environment import QuadraticFunction
from .synthetic_adaptive_environment import SynAdaptiveEnv from .synthetic_adaptive_environment import SynAdaptiveEnv

503
lib/datasets/get_dataset_with_transform.py
View File

@ -14,214 +14,349 @@ from .SearchDatasetWrap import SearchDataset
from config_utils import load_config from config_utils import load_config
Dataset2Class = {'cifar10' : 10, Dataset2Class = {
'cifar100': 100, "cifar10": 10,
'imagenet-1k-s':1000, "cifar100": 100,
'imagenet-1k' : 1000, "imagenet-1k-s": 1000,
'ImageNet16' : 1000, "imagenet-1k": 1000,
'ImageNet16-150': 150, "ImageNet16": 1000,
'ImageNet16-120': 120, "ImageNet16-150": 150,
'ImageNet16-200': 200} "ImageNet16-120": 120,
"ImageNet16-200": 200,
}
class CUTOUT(object): class CUTOUT(object):
def __init__(self, length):
self.length = length
def __init__(self, length): def __repr__(self):
self.length = length return "{name}(length={length})".format(
name=self.__class__.__name__, **self.__dict__
)
def __repr__(self): def __call__(self, img):
return ('{name}(length={length})'.format(name=self.__class__.__name__, **self.__dict__)) h, w = img.size(1), img.size(2)
mask = np.ones((h, w), np.float32)
y = np.random.randint(h)
x = np.random.randint(w)
def __call__(self, img): y1 = np.clip(y - self.length // 2, 0, h)
h, w = img.size(1), img.size(2) y2 = np.clip(y + self.length // 2, 0, h)
mask = np.ones((h, w), np.float32) x1 = np.clip(x - self.length // 2, 0, w)
y = np.random.randint(h) x2 = np.clip(x + self.length // 2, 0, w)
x = np.random.randint(w)
y1 = np.clip(y - self.length // 2, 0, h) mask[y1:y2, x1:x2] = 0.0
y2 = np.clip(y + self.length // 2, 0, h) mask = torch.from_numpy(mask)
x1 = np.clip(x - self.length // 2, 0, w) mask = mask.expand_as(img)
x2 = np.clip(x + self.length // 2, 0, w) img *= mask
return img
mask[y1: y2, x1: x2] = 0.
mask = torch.from_numpy(mask)
mask = mask.expand_as(img)
img *= mask
return img
imagenet_pca = { imagenet_pca = {
'eigval': np.asarray([0.2175, 0.0188, 0.0045]), "eigval": np.asarray([0.2175, 0.0188, 0.0045]),
'eigvec': np.asarray([ "eigvec": np.asarray(
[-0.5675, 0.7192, 0.4009], [
[-0.5808, -0.0045, -0.8140], [-0.5675, 0.7192, 0.4009],
[-0.5836, -0.6948, 0.4203], [-0.5808, -0.0045, -0.8140],
]) [-0.5836, -0.6948, 0.4203],
]
),
} }
class Lighting(object): class Lighting(object):
def __init__(self, alphastd, def __init__(
eigval=imagenet_pca['eigval'], self, alphastd, eigval=imagenet_pca["eigval"], eigvec=imagenet_pca["eigvec"]
eigvec=imagenet_pca['eigvec']): ):
self.alphastd = alphastd self.alphastd = alphastd
assert eigval.shape == (3,) assert eigval.shape == (3,)
assert eigvec.shape == (3, 3) assert eigvec.shape == (3, 3)
self.eigval = eigval self.eigval = eigval
self.eigvec = eigvec self.eigvec = eigvec
def __call__(self, img): def __call__(self, img):
if self.alphastd == 0.: if self.alphastd == 0.0:
return img return img
rnd = np.random.randn(3) * self.alphastd rnd = np.random.randn(3) * self.alphastd
rnd = rnd.astype('float32') rnd = rnd.astype("float32")
v = rnd v = rnd
old_dtype = np.asarray(img).dtype old_dtype = np.asarray(img).dtype
v = v * self.eigval v = v * self.eigval
v = v.reshape((3, 1)) v = v.reshape((3, 1))
inc = np.dot(self.eigvec, v).reshape((3,)) inc = np.dot(self.eigvec, v).reshape((3,))
img = np.add(img, inc) img = np.add(img, inc)
if old_dtype == np.uint8: if old_dtype == np.uint8:
img = np.clip(img, 0, 255) img = np.clip(img, 0, 255)
img = Image.fromarray(img.astype(old_dtype), 'RGB') img = Image.fromarray(img.astype(old_dtype), "RGB")
return img return img
def __repr__(self): def __repr__(self):
return self.__class__.__name__ + '()' return self.__class__.__name__ + "()"
def get_datasets(name, root, cutout): def get_datasets(name, root, cutout):
if name == 'cifar10': if name == "cifar10":
mean = [x / 255 for x in [125.3, 123.0, 113.9]] mean = [x / 255 for x in [125.3, 123.0, 113.9]]
std = [x / 255 for x in [63.0, 62.1, 66.7]] std = [x / 255 for x in [63.0, 62.1, 66.7]]
elif name == 'cifar100': elif name == "cifar100":
mean = [x / 255 for x in [129.3, 124.1, 112.4]] mean = [x / 255 for x in [129.3, 124.1, 112.4]]
std = [x / 255 for x in [68.2, 65.4, 70.4]] std = [x / 255 for x in [68.2, 65.4, 70.4]]
elif name.startswith('imagenet-1k'): elif name.startswith("imagenet-1k"):
mean, std = [0.485, 0.456, 0.406], [0.229, 0.224, 0.225] mean, std = [0.485, 0.456, 0.406], [0.229, 0.224, 0.225]
elif name.startswith('ImageNet16'): elif name.startswith("ImageNet16"):
mean = [x / 255 for x in [122.68, 116.66, 104.01]] mean = [x / 255 for x in [122.68, 116.66, 104.01]]
std = [x / 255 for x in [63.22, 61.26 , 65.09]] std = [x / 255 for x in [63.22, 61.26, 65.09]]
else: else:
raise TypeError("Unknow dataset : {:}".format(name)) raise TypeError("Unknow dataset : {:}".format(name))
# Data Argumentation # Data Argumentation
if name == 'cifar10' or name == 'cifar100': if name == "cifar10" or name == "cifar100":
lists = [transforms.RandomHorizontalFlip(), transforms.RandomCrop(32, padding=4), transforms.ToTensor(), transforms.Normalize(mean, std)] lists = [
if cutout > 0 : lists += [CUTOUT(cutout)] transforms.RandomHorizontalFlip(),
train_transform = transforms.Compose(lists) transforms.RandomCrop(32, padding=4),
test_transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize(mean, std)]) transforms.ToTensor(),
xshape = (1, 3, 32, 32) transforms.Normalize(mean, std),
elif name.startswith('ImageNet16'): ]
lists = [transforms.RandomHorizontalFlip(), transforms.RandomCrop(16, padding=2), transforms.ToTensor(), transforms.Normalize(mean, std)] if cutout > 0:
if cutout > 0 : lists += [CUTOUT(cutout)] lists += [CUTOUT(cutout)]
train_transform = transforms.Compose(lists) train_transform = transforms.Compose(lists)
test_transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize(mean, std)]) test_transform = transforms.Compose(
xshape = (1, 3, 16, 16) [transforms.ToTensor(), transforms.Normalize(mean, std)]
elif name == 'tiered': )
lists = [transforms.RandomHorizontalFlip(), transforms.RandomCrop(80, padding=4), transforms.ToTensor(), transforms.Normalize(mean, std)] xshape = (1, 3, 32, 32)
if cutout > 0 : lists += [CUTOUT(cutout)] elif name.startswith("ImageNet16"):
train_transform = transforms.Compose(lists) lists = [
test_transform = transforms.Compose([transforms.CenterCrop(80), transforms.ToTensor(), transforms.Normalize(mean, std)]) transforms.RandomHorizontalFlip(),
xshape = (1, 3, 32, 32) transforms.RandomCrop(16, padding=2),
elif name.startswith('imagenet-1k'): transforms.ToTensor(),
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) transforms.Normalize(mean, std),
if name == 'imagenet-1k': ]
xlists = [transforms.RandomResizedCrop(224)] if cutout > 0:
xlists.append( lists += [CUTOUT(cutout)]
transforms.ColorJitter( train_transform = transforms.Compose(lists)
brightness=0.4, test_transform = transforms.Compose(
contrast=0.4, [transforms.ToTensor(), transforms.Normalize(mean, std)]
saturation=0.4, )
hue=0.2)) xshape = (1, 3, 16, 16)
xlists.append( Lighting(0.1)) elif name == "tiered":
elif name == 'imagenet-1k-s': lists = [
xlists = [transforms.RandomResizedCrop(224, scale=(0.2, 1.0))] transforms.RandomHorizontalFlip(),
else: raise ValueError('invalid name : {:}'.format(name)) transforms.RandomCrop(80, padding=4),
xlists.append( transforms.RandomHorizontalFlip(p=0.5) ) transforms.ToTensor(),
xlists.append( transforms.ToTensor() ) transforms.Normalize(mean, std),
xlists.append( normalize ) ]
train_transform = transforms.Compose(xlists) if cutout > 0:
test_transform = transforms.Compose([transforms.Resize(256), transforms.CenterCrop(224), transforms.ToTensor(), normalize]) lists += [CUTOUT(cutout)]
xshape = (1, 3, 224, 224) train_transform = transforms.Compose(lists)
else: test_transform = transforms.Compose(
raise TypeError("Unknow dataset : {:}".format(name)) [
transforms.CenterCrop(80),
transforms.ToTensor(),
transforms.Normalize(mean, std),
]
)
xshape = (1, 3, 32, 32)
elif name.startswith("imagenet-1k"):
normalize = transforms.Normalize(
mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
)
if name == "imagenet-1k":
xlists = [transforms.RandomResizedCrop(224)]
xlists.append(
transforms.ColorJitter(
brightness=0.4, contrast=0.4, saturation=0.4, hue=0.2
)
)
xlists.append(Lighting(0.1))
elif name == "imagenet-1k-s":
xlists = [transforms.RandomResizedCrop(224, scale=(0.2, 1.0))]
else:
raise ValueError("invalid name : {:}".format(name))
xlists.append(transforms.RandomHorizontalFlip(p=0.5))
xlists.append(transforms.ToTensor())
xlists.append(normalize)
train_transform = transforms.Compose(xlists)
test_transform = transforms.Compose(
[
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
normalize,
]
)
xshape = (1, 3, 224, 224)
else:
raise TypeError("Unknow dataset : {:}".format(name))
if name == 'cifar10': if name == "cifar10":
train_data = dset.CIFAR10 (root, train=True , transform=train_transform, download=True) train_data = dset.CIFAR10(
test_data = dset.CIFAR10 (root, train=False, transform=test_transform , download=True) root, train=True, transform=train_transform, download=True
assert len(train_data) == 50000 and len(test_data) == 10000 )
elif name == 'cifar100': test_data = dset.CIFAR10(
train_data = dset.CIFAR100(root, train=True , transform=train_transform, download=True) root, train=False, transform=test_transform, download=True
test_data = dset.CIFAR100(root, train=False, transform=test_transform , download=True) )
assert len(train_data) == 50000 and len(test_data) == 10000 assert len(train_data) == 50000 and len(test_data) == 10000
elif name.startswith('imagenet-1k'): elif name == "cifar100":
train_data = dset.ImageFolder(osp.join(root, 'train'), train_transform) train_data = dset.CIFAR100(
test_data = dset.ImageFolder(osp.join(root, 'val'), test_transform) root, train=True, transform=train_transform, download=True
assert len(train_data) == 1281167 and len(test_data) == 50000, 'invalid number of images : {:} & {:} vs {:} & {:}'.format(len(train_data), len(test_data), 1281167, 50000) )
elif name == 'ImageNet16': test_data = dset.CIFAR100(
train_data = ImageNet16(root, True , train_transform) root, train=False, transform=test_transform, download=True
test_data = ImageNet16(root, False, test_transform) )
assert len(train_data) == 1281167 and len(test_data) == 50000 assert len(train_data) == 50000 and len(test_data) == 10000
elif name == 'ImageNet16-120': elif name.startswith("imagenet-1k"):
train_data = ImageNet16(root, True , train_transform, 120) train_data = dset.ImageFolder(osp.join(root, "train"), train_transform)
test_data = ImageNet16(root, False, test_transform , 120) test_data = dset.ImageFolder(osp.join(root, "val"), test_transform)
assert len(train_data) == 151700 and len(test_data) == 6000 assert (
elif name == 'ImageNet16-150': len(train_data) == 1281167 and len(test_data) == 50000
train_data = ImageNet16(root, True , train_transform, 150) ), "invalid number of images : {:} & {:} vs {:} & {:}".format(
test_data = ImageNet16(root, False, test_transform , 150) len(train_data), len(test_data), 1281167, 50000
assert len(train_data) == 190272 and len(test_data) == 7500 )
elif name == 'ImageNet16-200': elif name == "ImageNet16":
train_data = ImageNet16(root, True , train_transform, 200) train_data = ImageNet16(root, True, train_transform)
test_data = ImageNet16(root, False, test_transform , 200) test_data = ImageNet16(root, False, test_transform)
assert len(train_data) == 254775 and len(test_data) == 10000 assert len(train_data) == 1281167 and len(test_data) == 50000
else: raise TypeError("Unknow dataset : {:}".format(name)) elif name == "ImageNet16-120":
train_data = ImageNet16(root, True, train_transform, 120)
class_num = Dataset2Class[name] test_data = ImageNet16(root, False, test_transform, 120)
return train_data, test_data, xshape, class_num assert len(train_data) == 151700 and len(test_data) == 6000
elif name == "ImageNet16-150":
train_data = ImageNet16(root, True, train_transform, 150)
test_data = ImageNet16(root, False, test_transform, 150)
assert len(train_data) == 190272 and len(test_data) == 7500
elif name == "ImageNet16-200":
train_data = ImageNet16(root, True, train_transform, 200)
test_data = ImageNet16(root, False, test_transform, 200)
assert len(train_data) == 254775 and len(test_data) == 10000
else:
raise TypeError("Unknow dataset : {:}".format(name))
class_num = Dataset2Class[name]
return train_data, test_data, xshape, class_num
def get_nas_search_loaders(train_data, valid_data, dataset, config_root, batch_size, workers): def get_nas_search_loaders(
if isinstance(batch_size, (list,tuple)): train_data, valid_data, dataset, config_root, batch_size, workers
batch, test_batch = batch_size ):
else: if isinstance(batch_size, (list, tuple)):
batch, test_batch = batch_size, batch_size batch, test_batch = batch_size
if dataset == 'cifar10': else:
#split_Fpath = 'configs/nas-benchmark/cifar-split.txt' batch, test_batch = batch_size, batch_size
cifar_split = load_config('{:}/cifar-split.txt'.format(config_root), None, None) if dataset == "cifar10":
train_split, valid_split = cifar_split.train, cifar_split.valid # search over the proposed training and validation set # split_Fpath = 'configs/nas-benchmark/cifar-split.txt'
#logger.log('Load split file from {:}'.format(split_Fpath)) # they are two disjoint groups in the original CIFAR-10 training set cifar_split = load_config("{:}/cifar-split.txt".format(config_root), None, None)
# To split data train_split, valid_split = (
xvalid_data = deepcopy(train_data) cifar_split.train,
if hasattr(xvalid_data, 'transforms'): # to avoid a print issue cifar_split.valid,
xvalid_data.transforms = valid_data.transform ) # search over the proposed training and validation set
xvalid_data.transform = deepcopy( valid_data.transform ) # logger.log('Load split file from {:}'.format(split_Fpath)) # they are two disjoint groups in the original CIFAR-10 training set
search_data = SearchDataset(dataset, train_data, train_split, valid_split) # To split data
# data loader xvalid_data = deepcopy(train_data)
search_loader = torch.utils.data.DataLoader(search_data, batch_size=batch, shuffle=True , num_workers=workers, pin_memory=True) if hasattr(xvalid_data, "transforms"): # to avoid a print issue
train_loader = torch.utils.data.DataLoader(train_data , batch_size=batch, sampler=torch.utils.data.sampler.SubsetRandomSampler(train_split), num_workers=workers, pin_memory=True) xvalid_data.transforms = valid_data.transform
valid_loader = torch.utils.data.DataLoader(xvalid_data, batch_size=test_batch, sampler=torch.utils.data.sampler.SubsetRandomSampler(valid_split), num_workers=workers, pin_memory=True) xvalid_data.transform = deepcopy(valid_data.transform)
elif dataset == 'cifar100': search_data = SearchDataset(dataset, train_data, train_split, valid_split)
cifar100_test_split = load_config('{:}/cifar100-test-split.txt'.format(config_root), None, None) # data loader
search_train_data = train_data search_loader = torch.utils.data.DataLoader(
search_valid_data = deepcopy(valid_data) ; search_valid_data.transform = train_data.transform search_data,
search_data = SearchDataset(dataset, [search_train_data,search_valid_data], list(range(len(search_train_data))), cifar100_test_split.xvalid) batch_size=batch,
search_loader = torch.utils.data.DataLoader(search_data, batch_size=batch, shuffle=True , num_workers=workers, pin_memory=True) shuffle=True,
train_loader = torch.utils.data.DataLoader(train_data , batch_size=batch, shuffle=True , num_workers=workers, pin_memory=True) num_workers=workers,
valid_loader = torch.utils.data.DataLoader(valid_data , batch_size=test_batch, sampler=torch.utils.data.sampler.SubsetRandomSampler(cifar100_test_split.xvalid), num_workers=workers, pin_memory=True) pin_memory=True,
elif dataset == 'ImageNet16-120': )
imagenet_test_split = load_config('{:}/imagenet-16-120-test-split.txt'.format(config_root), None, None) train_loader = torch.utils.data.DataLoader(
search_train_data = train_data train_data,
search_valid_data = deepcopy(valid_data) ; search_valid_data.transform = train_data.transform batch_size=batch,
search_data = SearchDataset(dataset, [search_train_data,search_valid_data], list(range(len(search_train_data))), imagenet_test_split.xvalid) sampler=torch.utils.data.sampler.SubsetRandomSampler(train_split),
search_loader = torch.utils.data.DataLoader(search_data, batch_size=batch, shuffle=True , num_workers=workers, pin_memory=True) num_workers=workers,
train_loader = torch.utils.data.DataLoader(train_data , batch_size=batch, shuffle=True , num_workers=workers, pin_memory=True) pin_memory=True,
valid_loader = torch.utils.data.DataLoader(valid_data , batch_size=test_batch, sampler=torch.utils.data.sampler.SubsetRandomSampler(imagenet_test_split.xvalid), num_workers=workers, pin_memory=True) )
else: valid_loader = torch.utils.data.DataLoader(
raise ValueError('invalid dataset : {:}'.format(dataset)) xvalid_data,
return search_loader, train_loader, valid_loader batch_size=test_batch,
sampler=torch.utils.data.sampler.SubsetRandomSampler(valid_split),
num_workers=workers,
pin_memory=True,
)
elif dataset == "cifar100":
cifar100_test_split = load_config(
"{:}/cifar100-test-split.txt".format(config_root), None, None
)
search_train_data = train_data
search_valid_data = deepcopy(valid_data)
search_valid_data.transform = train_data.transform
search_data = SearchDataset(
dataset,
[search_train_data, search_valid_data],
list(range(len(search_train_data))),
cifar100_test_split.xvalid,
)
search_loader = torch.utils.data.DataLoader(
search_data,
batch_size=batch,
shuffle=True,
num_workers=workers,
pin_memory=True,
)
train_loader = torch.utils.data.DataLoader(
train_data,
batch_size=batch,
shuffle=True,
num_workers=workers,
pin_memory=True,
)
valid_loader = torch.utils.data.DataLoader(
valid_data,
batch_size=test_batch,
sampler=torch.utils.data.sampler.SubsetRandomSampler(
cifar100_test_split.xvalid
),
num_workers=workers,
pin_memory=True,
)
elif dataset == "ImageNet16-120":
imagenet_test_split = load_config(
"{:}/imagenet-16-120-test-split.txt".format(config_root), None, None
)
search_train_data = train_data
search_valid_data = deepcopy(valid_data)
search_valid_data.transform = train_data.transform
search_data = SearchDataset(
dataset,
[search_train_data, search_valid_data],
list(range(len(search_train_data))),
imagenet_test_split.xvalid,
)
search_loader = torch.utils.data.DataLoader(
search_data,
batch_size=batch,
shuffle=True,
num_workers=workers,
pin_memory=True,
)
train_loader = torch.utils.data.DataLoader(
train_data,
batch_size=batch,
shuffle=True,
num_workers=workers,
pin_memory=True,
)
valid_loader = torch.utils.data.DataLoader(
valid_data,
batch_size=test_batch,
sampler=torch.utils.data.sampler.SubsetRandomSampler(
imagenet_test_split.xvalid
),
num_workers=workers,
pin_memory=True,
)
else:
raise ValueError("invalid dataset : {:}".format(dataset))
return search_loader, train_loader, valid_loader
#if __name__ == '__main__':
# if __name__ == '__main__':
# train_data, test_data, xshape, class_num = dataset = get_datasets('cifar10', '/data02/dongxuanyi/.torch/cifar.python/', -1) # train_data, test_data, xshape, class_num = dataset = get_datasets('cifar10', '/data02/dongxuanyi/.torch/cifar.python/', -1)
# import pdb; pdb.set_trace() # import pdb; pdb.set_trace()

273
lib/datasets/landmark_utils/point_meta.py
View File

@ -9,108 +9,211 @@ from xvision import normalize_points
from xvision import denormalize_points from xvision import denormalize_points
class PointMeta(): class PointMeta:
# points : 3 x num_pts (x, y, oculusion) # points : 3 x num_pts (x, y, oculusion)
# image_size: original [width, height] # image_size: original [width, height]
def __init__(self, num_point, points, box, image_path, dataset_name): def __init__(self, num_point, points, box, image_path, dataset_name):
self.num_point = num_point self.num_point = num_point
if box is not None: if box is not None:
assert (isinstance(box, tuple) or isinstance(box, list)) and len(box) == 4 assert (isinstance(box, tuple) or isinstance(box, list)) and len(box) == 4
self.box = torch.Tensor(box) self.box = torch.Tensor(box)
else: self.box = None else:
if points is None: self.box = None
self.points = points if points is None:
else: self.points = points
assert len(points.shape) == 2 and points.shape[0] == 3 and points.shape[1] == self.num_point, 'The shape of point is not right : {}'.format( points ) else:
self.points = torch.Tensor(points.copy()) assert (
self.image_path = image_path len(points.shape) == 2
self.datasets = dataset_name and points.shape[0] == 3
and points.shape[1] == self.num_point
), "The shape of point is not right : {}".format(points)
self.points = torch.Tensor(points.copy())
self.image_path = image_path
self.datasets = dataset_name
def __repr__(self): def __repr__(self):
if self.box is None: boxstr = 'None' if self.box is None:
else : boxstr = 'box=[{:.1f}, {:.1f}, {:.1f}, {:.1f}]'.format(*self.box.tolist()) boxstr = "None"
return ('{name}(points={num_point}, '.format(name=self.__class__.__name__, **self.__dict__) + boxstr + ')') else:
boxstr = "box=[{:.1f}, {:.1f}, {:.1f}, {:.1f}]".format(*self.box.tolist())
return (
"{name}(points={num_point}, ".format(
name=self.__class__.__name__, **self.__dict__
)
+ boxstr
+ ")"
)
def get_box(self, return_diagonal=False): def get_box(self, return_diagonal=False):
if self.box is None: return None if self.box is None:
if not return_diagonal: return None
return self.box.clone() if not return_diagonal:
else: return self.box.clone()
W = (self.box[2]-self.box[0]).item() else:
H = (self.box[3]-self.box[1]).item() W = (self.box[2] - self.box[0]).item()
return math.sqrt(H*H+W*W) H = (self.box[3] - self.box[1]).item()
return math.sqrt(H * H + W * W)
def get_points(self, ignore_indicator=False): def get_points(self, ignore_indicator=False):
if ignore_indicator: last = 2 if ignore_indicator:
else : last = 3 last = 2
if self.points is not None: return self.points.clone()[:last, :] else:
else : return torch.zeros((last, self.num_point)) last = 3
if self.points is not None:
return self.points.clone()[:last, :]
else:
return torch.zeros((last, self.num_point))
def is_none(self): def is_none(self):
#assert self.box is not None, 'The box should not be None' # assert self.box is not None, 'The box should not be None'
return self.points is None return self.points is None
#if self.box is None: return True # if self.box is None: return True
#else : return self.points is None # else : return self.points is None
def copy(self): def copy(self):
return copy.deepcopy(self) return copy.deepcopy(self)
def visiable_pts_num(self): def visiable_pts_num(self):
with torch.no_grad(): with torch.no_grad():
ans = self.points[2,:] > 0 ans = self.points[2, :] > 0
ans = torch.sum(ans) ans = torch.sum(ans)
ans = ans.item() ans = ans.item()
return ans return ans
def special_fun(self, indicator):
if indicator == '68to49': # For 300W or 300VW, convert the default 68 points to 49 points.
assert self.num_point == 68, 'num-point must be 68 vs. {:}'.format(self.num_point)
self.num_point = 49
out = torch.ones((68), dtype=torch.uint8)
out[[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,60,64]] = 0
if self.points is not None: self.points = self.points.clone()[:, out]
else:
raise ValueError('Invalid indicator : {:}'.format( indicator ))
def apply_horizontal_flip(self): def special_fun(self, indicator):
#self.points[0, :] = width - self.points[0, :] - 1 if (
# Mugsy spefic or Synthetic indicator == "68to49"
if self.datasets.startswith('HandsyROT'): ): # For 300W or 300VW, convert the default 68 points to 49 points.
ori = np.array(list(range(0, 42))) assert self.num_point == 68, "num-point must be 68 vs. {:}".format(
pos = np.array(list(range(21,42)) + list(range(0,21))) self.num_point
self.points[:, pos] = self.points[:, ori] )
elif self.datasets.startswith('face68'): self.num_point = 49
ori = np.array(list(range(0, 68))) out = torch.ones((68), dtype=torch.uint8)
pos = np.array([17,16,15,14,13,12,11,10, 9, 8,7,6,5,4,3,2,1, 27,26,25,24,23,22,21,20,19,18, 28,29,30,31, 36,35,34,33,32, 46,45,44,43,48,47, 40,39,38,37,42,41, 55,54,53,52,51,50,49,60,59,58,57,56,65,64,63,62,61,68,67,66])-1 out[[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 60, 64]] = 0
self.points[:, ori] = self.points[:, pos] if self.points is not None:
else: self.points = self.points.clone()[:, out]
raise ValueError('Does not support {:}'.format(self.datasets)) else:
raise ValueError("Invalid indicator : {:}".format(indicator))
def apply_horizontal_flip(self):
# self.points[0, :] = width - self.points[0, :] - 1
# Mugsy spefic or Synthetic
if self.datasets.startswith("HandsyROT"):
ori = np.array(list(range(0, 42)))
pos = np.array(list(range(21, 42)) + list(range(0, 21)))
self.points[:, pos] = self.points[:, ori]
elif self.datasets.startswith("face68"):
ori = np.array(list(range(0, 68)))
pos = (
np.array(
[
17,
16,
15,
14,
13,
12,
11,
10,
9,
8,
7,
6,
5,
4,
3,
2,
1,
27,
26,
25,
24,
23,
22,
21,
20,
19,
18,
28,
29,
30,
31,
36,
35,
34,
33,
32,
46,
45,
44,
43,
48,
47,
40,
39,
38,
37,
42,
41,
55,
54,
53,
52,
51,
50,
49,
60,
59,
58,
57,
56,
65,
64,
63,
62,
61,
68,
67,
66,
]
)
- 1
)
self.points[:, ori] = self.points[:, pos]
else:
raise ValueError("Does not support {:}".format(self.datasets))
# shape = (H,W) # shape = (H,W)
def apply_affine2point(points, theta, shape): def apply_affine2point(points, theta, shape):
assert points.size(0) == 3, 'invalid points shape : {:}'.format(points.size()) assert points.size(0) == 3, "invalid points shape : {:}".format(points.size())
with torch.no_grad(): with torch.no_grad():
ok_points = points[2,:] == 1 ok_points = points[2, :] == 1
assert torch.sum(ok_points).item() > 0, 'there is no visiable point' assert torch.sum(ok_points).item() > 0, "there is no visiable point"
points[:2,:] = normalize_points(shape, points[:2,:]) points[:2, :] = normalize_points(shape, points[:2, :])
norm_trans_points = ok_points.unsqueeze(0).repeat(3, 1).float() norm_trans_points = ok_points.unsqueeze(0).repeat(3, 1).float()
trans_points, ___ = torch.gesv(points[:, ok_points], theta) trans_points, ___ = torch.gesv(points[:, ok_points], theta)
norm_trans_points[:, ok_points] = trans_points norm_trans_points[:, ok_points] = trans_points
return norm_trans_points
return norm_trans_points
def apply_boundary(norm_trans_points): def apply_boundary(norm_trans_points):
with torch.no_grad(): with torch.no_grad():
norm_trans_points = norm_trans_points.clone() norm_trans_points = norm_trans_points.clone()
oks = torch.stack((norm_trans_points[0]>-1, norm_trans_points[0]<1, norm_trans_points[1]>-1, norm_trans_points[1]<1, norm_trans_points[2]>0)) oks = torch.stack(
oks = torch.sum(oks, dim=0) == 5 (
norm_trans_points[2, :] = oks norm_trans_points[0] > -1,
return norm_trans_points norm_trans_points[0] < 1,
norm_trans_points[1] > -1,
norm_trans_points[1] < 1,
norm_trans_points[2] > 0,
)
)
oks = torch.sum(oks, dim=0) == 5
norm_trans_points[2, :] = oks
return norm_trans_points

133
lib/datasets/synthetic_adaptive_environment.py
View File

@ -1,39 +1,123 @@
##################################################### #####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.03 # # Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.03 #
##################################################### #####################################################
import math
import numpy as np import numpy as np
from typing import Optional from typing import Optional
import torch
import torch.utils.data as data import torch.utils.data as data
class QuadraticFunction:
"""The quadratic function that outputs f(x) = a * x^2 + b * x + c."""
def __init__(self, list_of_points=None):
self._params = dict(a=None, b=None, c=None)
if list_of_points is not None:
self.fit(list_of_points)
def set(self, a, b, c):
self._params["a"] = a
self._params["b"] = b
self._params["c"] = c
def check_valid(self):
for key, value in self._params.items():
if value is None:
raise ValueError("The {:} is None".format(key))
def __getitem__(self, x):
self.check_valid()
return self._params["a"] * x * x + self._params["b"] * x + self._params["c"]
def fit(
self,
list_of_points,
transf=lambda x: x,
max_iter=900,
lr_max=1.0,
verbose=False,
):
with torch.no_grad():
data = torch.Tensor(list_of_points).type(torch.float32)
assert data.ndim == 2 and data.size(1) == 2, "Invalid shape : {:}".format(
data.shape
)
x, y = data[:, 0], data[:, 1]
weights = torch.nn.Parameter(torch.Tensor(3))
torch.nn.init.normal_(weights, mean=0.0, std=1.0)
optimizer = torch.optim.Adam([weights], lr=lr_max, amsgrad=True)
lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=[int(max_iter*0.25), int(max_iter*0.5), int(max_iter*0.75)], gamma=0.1)
if verbose:
print("The optimizer: {:}".format(optimizer))
best_loss = None
for _iter in range(max_iter):
y_hat = transf(weights[0] * x * x + weights[1] * x + weights[2])
loss = torch.mean(torch.abs(y - y_hat))
optimizer.zero_grad()
loss.backward()
optimizer.step()
lr_scheduler.step()
if verbose:
print(
"In QuadraticFunction's fit, loss at the {:02d}/{:02d}-th iter is {:}".format(
_iter, max_iter, loss.item()
)
)
# Update the params
if best_loss is None or best_loss > loss.item():
best_loss = loss.item()
self._params["a"] = weights[0].item()
self._params["b"] = weights[1].item()
self._params["c"] = weights[2].item()
def __repr__(self):
return "{name}(y = {a} * x^2 + {b} * x + {c})".format(
name=self.__class__.__name__,
a=self._params["a"],
b=self._params["b"],
c=self._params["c"],
)
class SynAdaptiveEnv(data.Dataset): class SynAdaptiveEnv(data.Dataset):
"""The synethtic dataset for adaptive environment.""" """The synethtic dataset for adaptive environment.
- x in [0, 1]
- y = amplitude-scale-of(x) * sin( period-phase-shift-of(x) )
- where
- the amplitude scale is a quadratic function of x
- the period-phase-shift is another quadratic function of x
"""
def __init__( def __init__(
self, self,
max_num_phase: int = 100, num: int = 100,
interval: float = 0.1, num_sin_phase: int = 4,
max_scale: float = 4, min_amplitude: float = 1,
offset_scale: float = 1.5, max_amplitude: float = 4,
phase_shift: float = 0,
mode: Optional[str] = None, mode: Optional[str] = None,
): ):
self._amplitude_scale = QuadraticFunction(
[(0, min_amplitude), (0.5, max_amplitude), (0, min_amplitude)]
)
self._max_num_phase = max_num_phase self._num_sin_phase = num_sin_phase
self._interval = interval self._interval = 1.0 / (float(num) - 1)
self._total_num = num
self._period_phase_shift = QuadraticFunction()
fitting_data = []
temp_max_scalar = 2 ** num_sin_phase
for i in range(num_sin_phase):
value = (2 ** i) / temp_max_scalar
fitting_data.append((value, math.sin(value)))
self._period_phase_shift.fit(fitting_data, transf=lambda x: torch.sin(x))
self._times = np.arange(0, np.pi * self._max_num_phase, self._interval)
xmin, xmax = self._times.min(), self._times.max()
self._inputs = []
self._total_num = len(self._times)
for i in range(self._total_num):
scale = (i + 1.0) / self._total_num * max_scale
sin_scale = (i + 1.0) / self._total_num * 0.7
sin_scale = -4 * (sin_scale - 0.5) ** 2 + 1
# scale = -(self._times[i] - (xmin - xmax) / 2) + max_scale
self._inputs.append(
np.sin(self._times[i] * sin_scale) * (offset_scale - scale)
)
self._inputs = np.array(self._inputs)
# Training Set 60% # Training Set 60%
num_of_train = int(self._total_num * 0.6) num_of_train = int(self._total_num * 0.6)
# Validation Set 20% # Validation Set 20%
@ -70,10 +154,11 @@ class SynAdaptiveEnv(data.Dataset):
def __getitem__(self, index): def __getitem__(self, index):
assert 0 <= index < len(self), "{:} is not in [0, {:})".format(index, len(self)) assert 0 <= index < len(self), "{:} is not in [0, {:})".format(index, len(self))
index = self._indexes[index] index = self._indexes[index]
value = float(self._inputs[index]) position = self._interval * index
if self._transform is not None: value = self._amplitude_scale[position] * math.sin(
value = self._transform(value) self._period_phase_shift[position]
return index, float(self._times[index]), value )
return index, position, value
def __len__(self): def __len__(self):
return len(self._indexes) return len(self._indexes)

30
lib/datasets/test_utils.py
View File

@ -5,16 +5,20 @@ import os
def test_imagenet_data(imagenet): def test_imagenet_data(imagenet):
total_length = len(imagenet) total_length = len(imagenet)
assert total_length == 1281166 or total_length == 50000, 'The length of ImageNet is wrong : {}'.format(total_length) assert (
map_id = {} total_length == 1281166 or total_length == 50000
for index in range(total_length): ), "The length of ImageNet is wrong : {}".format(total_length)
path, target = imagenet.imgs[index] map_id = {}
folder, image_name = os.path.split(path) for index in range(total_length):
_, folder = os.path.split(folder) path, target = imagenet.imgs[index]
if folder not in map_id: folder, image_name = os.path.split(path)
map_id[folder] = target _, folder = os.path.split(folder)
else: if folder not in map_id:
assert map_id[folder] == target, 'Class : {} is not {}'.format(folder, target) map_id[folder] = target
assert image_name.find(folder) == 0, '{} is wrong.'.format(path) else:
print ('Check ImageNet Dataset OK') assert map_id[folder] == target, "Class : {} is not {}".format(
folder, target
)
assert image_name.find(folder) == 0, "{} is wrong.".format(path)
print("Check ImageNet Dataset OK")

46
notebooks/TOT/synthetic.ipynb
View File

File diff suppressed because one or more lines are too long

24
tests/test_synthetic.py
View File

@ -13,9 +13,33 @@ print("library path: {:}".format(lib_dir))
if str(lib_dir) not in sys.path: if str(lib_dir) not in sys.path:
sys.path.insert(0, str(lib_dir)) sys.path.insert(0, str(lib_dir))
from datasets import QuadraticFunction
from datasets import SynAdaptiveEnv from datasets import SynAdaptiveEnv
class TestQuadraticFunction(unittest.TestCase):
"""Test the quadratic function."""
def test_simple(self):
function = QuadraticFunction([[0, 1], [0.5, 4], [1, 1]])
print(function)
for x in (0, 0.5, 1):
print("f({:})={:}".format(x, function[x]))
thresh = 0.2
self.assertTrue(abs(function[0] - 1) < thresh)
self.assertTrue(abs(function[0.5] - 4) < thresh)
self.assertTrue(abs(function[1] - 1) < thresh)
def test_none(self):
function = QuadraticFunction()
function.fit([[0, 1], [0.5, 4], [1, 1]], max_iter=3000, verbose=True)
print(function)
thresh = 0.2
self.assertTrue(abs(function[0] - 1) < thresh)
self.assertTrue(abs(function[0.5] - 4) < thresh)
self.assertTrue(abs(function[1] - 1) < thresh)
class TestSynAdaptiveEnv(unittest.TestCase): class TestSynAdaptiveEnv(unittest.TestCase):
"""Test the synethtic adaptive environment.""" """Test the synethtic adaptive environment."""