def initialize(self, source, target, batch_size1, batch_size2, scale=32):
transform = transforms.Compose([
transforms.Scale(scale),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
dataset_source = Dataset(source['imgs'],
source['labels'],
transform=transform)
dataset_target = Dataset(target['imgs'],
target['labels'],
transform=transform)
# dataset_source = tnt.dataset.TensorDataset([source['imgs'], source['labels']])
# dataset_target = tnt.dataset.TensorDataset([target['imgs'], target['labels']])
data_loader_s = torch.utils.data.DataLoader(dataset_source,
batch_size=batch_size1,
shuffle=True,
num_workers=4)
data_loader_t = torch.utils.data.DataLoader(dataset_target,
batch_size=batch_size2,
shuffle=True,
num_workers=4)
self.dataset_s = dataset_source
self.dataset_t = dataset_target
self.paired_data = PairedData(data_loader_s, data_loader_t,
float("inf"))
def test_check_hashes(self):
d = Dataset('fake_dataset', data_store='', data_root=self.data_root,
local_dir='/Users/fpbatta/local_store')
temp_dir = d.make_local_copy()
self.temp_dirs.append(temp_dir)
d.create_file_hashes()
self.assertTrue(d.check_file_hashes())
def test_subdirs(self):
d = Dataset('fake_dataset', data_store='', data_root=self.data_root,
local_dir='/Users/fpbatta/local_store', subdirs_as_datasets=True)
self.assertTrue(d.children)
temp_dir = d.make_local_copy()
self.temp_dirs.append(temp_dir)
self.assertEqual(os.listdir(temp_dir), ['a1.avi', 'fd1.dat', 'fd2.dat', 'fd3.dat'])
def test_temp_copy(self):
d = Dataset('fake_dataset', data_store='', data_root=self.data_root,
local_dir='/Users/fpbatta/local_store')
temp_dir = d.make_local_copy()
self.temp_dirs.append(temp_dir)
self.assertEqual(os.listdir(temp_dir), ['a1.avi', 'fd1.dat', 'fd2.dat', 'fd3.dat',
'the_subdir'])
def test_subdirs_temp_copy(self):
d = Dataset('fake_dataset', data_store='', data_root=self.data_root,
local_dir='/Users/fpbatta/local_store')
temp_dir = d.make_local_copy()
self.temp_dirs.append(temp_dir)
self.assertEqual(os.listdir(temp_dir), ['a1.avi', 'fd1.dat', 'fd2.dat', 'fd3.dat',
'the_subdir'])
self.assertTrue(os.path.isdir(os.path.join(temp_dir, 'the_subdir')))
logging.debug(os.listdir(os.path.join(temp_dir, 'the_subdir')))
def test_hashes(self):
d = Dataset('fake_dataset', data_store='', data_root=self.data_root,
local_dir='/Users/fpbatta/local_store')
temp_dir = d.make_local_copy()
self.temp_dirs.append(temp_dir)
d.create_file_hashes()
self.assertTrue(d.hashes)
h = hashlib.md5()
h.update(fd1_dat.encode('utf-8'))
self.assertEqual(h.hexdigest(), d.hashes['fd1.dat'])
h = hashlib.md5()
h.update(sfd1_dat.encode('utf-8'))
self.assertEqual(h.hexdigest(), d.hashes['the_subdir/sfd1.dat'])
logging.debug("hashes: " + str(d.hashes))
def initialize(self, source, target, batch_size1, batch_size2, scale=32):
# transform = transforms.Compose([
# transforms.Scale(scale),
# transforms.ToTensor(),
# transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
# ])
dataset_source = Dataset(source['imgs'], source['labels'])
dataset_target = Dataset(target['imgs'], target['labels'])
data_loader_s = torch.utils.data.DataLoader(dataset_source,
batch_size=batch_size1,
shuffle=True,
num_workers=0)
data_loader_t = torch.utils.data.DataLoader(dataset_target,
batch_size=batch_size2,
shuffle=True,
num_workers=0)
# print('Source shape: {}, target shape: {}'.format(len(data_loader_s), len(data_loader_t)))
self.dataset_s = dataset_source
self.dataset_t = dataset_target
self.paired_data = PairedData(data_loader_s, data_loader_t,
float("inf"))
if __name__ == "__main__":
from config import config
from datasets.datasets import Dataset
import matplotlib.patches as patches
import matplotlib.pyplot as plt
import torchvision.transforms as transforms
ii = r'/home/wei/Deep_learning_pytorch/Data/UCAS/ucas_train.txt'
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
transform = transforms.Compose([
transforms.ToTensor(),
])
img_size = 256
da = Dataset(ii, transform=transform, img_size=img_size, train=False)
dataloader = torch.utils.data.DataLoader(da, batch_size=1, shuffle=False)
#x = torch.randn(1,3,128,128)
f = FCOS(config)
#checkpoint = torch.load('./checkpoint/ckpt.pth')
#f.load_state_dict(checkpoint['weights'])
for batch_i, (_, imgs, targets) in enumerate(dataloader):
images = imgs
targets = targets
#loss = f(images, targets)
detections = f(images)
break
