def extract_all(fp, fmt=None, root='.'):
def find_decompressor(fmt_):
dc_map = {
'gzip': (tarfile.open, 'r:gz'),
'zip': (zipfile.ZipFile, 'r'),
'bz2': (tarfile.open, 'r:bz2')
}
if not fmt_ in dc_map:
raise ValueError('the `%s` format is not supported.' % (fmt_,))
return dc_map[fmt_]
if not fmt:
if fp.endswith('.zip'):
fmt = 'zip'
elif fp.endswith('.tar.gz') or fp.endswith('.tgz'):
fmt = 'gzip'
elif fp.endswith('.tar.bz2') or fp.endswith('.tbz'):
fmt = 'bz2'
else:
fmt = 'None'
opener, mode = find_decompressor(fmt)
try:
with opener(fp, mode) as f:
f.extractall(path=root)
except StandardError as e:
logger.fatal('error occur while extracting file %s: %s', fp, str(e))
def prompt_for_eula(self):
eula = self._eula_by_brand()
print eula
ret = self.io.require_confirmation("Do you accept the terms above?")
if not ret:
logger.fatal("Setup aborted, Cancelled by user")
quit()
def get_train_loader(conf, data_mode, sample_identity=False):
if data_mode == 'emore':
root = conf.emore_folder/'imgs'
elif data_mode == 'glint':
root = conf.glint_folder/'imgs'
else:
logger.fatal('invalide data_mode {}'.format(data_mode))
exit(1)
class_num, class_to_idx = find_classes(root)
train_transform = trans.Compose([
trans.RandomHorizontalFlip(),
trans.ColorJitter(brightness=0.2, contrast=0.15, saturation=0, hue=0),
trans.ToTensor(),
trans.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
])
extensions = ['.jpg', '.jpeg', '.png', '.ppm', '.bmp', '.pgm', '.tif']
path_ds = make_dataset(root, class_to_idx, extensions)
dataset = ImageDataset(path_ds, train_transform)
if sample_identity:
train_sampler = DistRandomIdentitySampler(dataset.dataset, conf.batch_size, conf.num_instances)
else:
train_sampler = distributed.DistributedSampler(dataset)
loader = DataLoader(dataset, batch_size=conf.batch_size, shuffle=False, pin_memory=conf.pin_memory, num_workers=conf.num_workers, sampler = train_sampler)
return loader, class_num
def prompt_for_eula(self):
eula = self._eula_by_brand()
print eula
ret = self.io.require_confirmation("Do you accept the terms above?")
if not ret:
logger.fatal("Setup aborted, Cancelled by user")
quit()
def main():
arguments = docopt.docopt(__doc__, version=VERSION_NUMBER)
input_map = arguments["<contact_map>"]
binning = int(arguments["--binning"])
normalized = arguments["--normalize"]
vmax = float(arguments["--max"])
output_file = arguments["--output"]
process_matrix = save_matrix
if not output_file or output_file == "output.png":
process_matrix = plot_matrix
raw_map = load_raw_matrix(input_map)
sparse_map = raw_cols_to_sparse(raw_map)
if normalized:
sparse_map = hcs.normalize_sparse(sparse_map, norm="SCN")
if binning > 1:
binned_map = hcs.bin_sparse(M=sparse_map, subsampling_factor=binning)
else:
binned_map = sparse_map
try:
dense_map = sparse_to_dense(binned_map)
process_matrix(dense_map, filename=output_file, vmax=vmax)
except MemoryError:
logger.fatal("Contact map is too large to load, try binning more")
def get_train_loader_from_txt(conf, data_mode, sample_identity=False):
if data_mode == 'emore':
txt_path = conf.emore_folder/'imgs'/'train_list.txt'
elif data_mode == 'glint':
txt_path = conf.glint_folder/'imgs'/'train_list.txt'
else:
logger.fatal('invalide data_mode {}'.format(data_mode))
exit(1)
train_transform = trans.Compose([
trans.RandomHorizontalFlip(),
trans.ColorJitter(brightness=0.2, contrast=0.15, saturation=0, hue=0),
trans.ToTensor(),
trans.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
])
dataset = ImageLandmarkDataset(txt_path, train_transform)
if sample_identity:
train_sampler = DistRandomIdentitySampler(dataset.dataset, conf.batch_size, conf.num_instances)
else:
train_sampler = distributed.DistributedSampler(dataset)
loader = DataLoader(dataset, batch_size=conf.batch_size, shuffle=False, pin_memory=conf.pin_memory, num_workers=conf.num_workers, sampler = train_sampler)
return loader, dataset.class_num
