def main():
from torchvision.utils import make_grid
from torchvision import transforms
from common.vis_utils import show_batch
import matplotlib.pyplot as plt
import os.path as osp
import os.path as osp
from maskrobotcar import RobotCar
from dataset_loader.dataloaders import get_mask_transforms, get_train_transforms
from common.utils import get_configuration
from mapnet.config import MapNetConfigurator
from dataloaders import get_mapnet_train_dataloader
config = get_configuration(MapNetConfigurator())
data_loader, val_dataloader = get_mapnet_train_dataloader(config)
# scene = 'full'
# num_workers = 4
# transform = transforms.Compose([
# transforms.Scale(256),
# transforms.CenterCrop(224),
# transforms.ToTensor()])
# mask_transform = get_mask_transforms(config)
# transform, t = get_train_transforms(config)
# data_path = osp.join('/home/drinkingcoder/Dataset/robotcar')
# data_dir = '../data/RobotCar/full/'
# dset = RobotCar(scene, data_path, train=True, real=True, mask_transform=mask_transform, transform=transform, data_dir=data_dir)
# print 'Loaded RobotCar scene {:s}, length = {:d}'.format(scene, len(dset))
# plot the poses
# plt.figure()
# plt.plot(dset.poses[:, 0], dset.poses[:, 1])
# plt.show()
# data_loader = data.DataLoader(dset, batch_size=10, shuffle=True,
# num_workers=num_workers)
batch_count = 0
N = 3
for batch in data_loader:
print 'Minibatch {:d}'.format(batch_count)
data, targ = batch
img, mask = data
print(img.size())
print(mask.size())
img = mask[0][0]
print(img.size())
print(torch.mean(mask[0][0]))
img = transforms.ToPILImage()(img)
# plt.imshow(batch[0][0, 3:])
img.show()
# show_batch(make_grid(img, nrow=5, padding=25, normalize=True))
# img = batch[0][1][0] * 255
# img = transforms.ToPILImage()(img)
# plt.imshow(batch[0][0, 3:])
# img.show()
batch_count += 1
if batch_count >= N:
break
def main():
from common.vis_utils import show_batch
from torchvision.utils import make_grid
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
from mapnet.config import MapNetConfigurator
from common.utils import get_configuration
scene = 'full'
num_workers = 4
transform = transforms.Compose([
transforms.Scale(256),
# transforms.CenterCrop(224),
transforms.ToTensor()])
config = get_configuration(MapNetConfigurator())
config.uniform_sampling = False
config.mask_sampling = False
config.mask_image = True
data_path = "/home/drinkingcoder/Dataset/robotcar/"
dset = RobotCar(scene, data_path, train=True, real=False,
transform=transform, data_dir=osp.join("..", "data", "RobotCar"),
config=config)
print 'Loaded RobotCar scene {:s}, length = {:d}'.format(scene, len(dset))
# plot the poses
plt.figure()
plt.plot(dset.poses[:, 0], dset.poses[:, 1])
plt.show()
print(len(dset))
data_loader = data.DataLoader(dset, batch_size=10, shuffle=True,
num_workers=num_workers)
#plt.imshow(dset.muimg.data.numpy()[0, :, :])
#plt.imshow(dset[0].data.numpy())
batch_count = 0
N = 2
for batch in data_loader:
print 'Minibatch {:d}'.format(batch_count)
show_batch(make_grid(batch[0], nrow=5, padding=25, normalize=True))
batch_count += 1
if batch_count >= N:
break
def main():
from common.vis_utils import show_batch
from torchvision.utils import make_grid
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
from mapnet.config import MapNetConfigurator
from common.utils import get_configuration
scene = 'KingsCollege'
num_workers = 4
transform = transforms.Compose([
transforms.Scale(256),
# transforms.CenterCrop(224),
transforms.ToTensor()])
config = get_configuration(MapNetConfigurator())
data_path = "/home/drinkingcoder/Dataset/Cambridge/"
dset = Cambridge(scene, data_path, train=True, real=False,
transform=transform, data_dir=osp.join("..", "data", "Cambridge", "KingsCollege"),
config=config)
print 'Loaded RobotCar scene {:s}, length = {:d}'.format(scene, len(dset))
# plot the poses
plt.figure()
plt.scatter(dset.poses[:, 0], dset.poses[:, 1])
plt.show()
plt.figure()
plt.scatter(dset.poses[:, 3], dset.poses[:, 5])
plt.show()
print(len(dset))
data_loader = data.DataLoader(dset, batch_size=10, shuffle=True,
num_workers=num_workers)
batch_count = 0
N = 2
for batch in data_loader:
print 'Minibatch {:d}'.format(batch_count)
show_batch(make_grid(batch[0], nrow=5, padding=25, normalize=True))
batch_count += 1
if batch_count >= N:
break
from mapnet.MapNet import MapNet
def get_args():
parser = argparse.ArgumentParser(description='Training script for mapnet')
parser.add_argument('--config', type=str, default=None)
args = parser.parse_args()
return args
if __name__ == "__main__":
# configuration
configuration = get_configuration(AttentionMapNetConfigurator(),
get_args())
# model
feature_extractor = models.resnet34(pretrained=True)
if configuration.model == "SEAttentionPoseNet":
posenet = SEAttentionPoseNet(resnet=feature_extractor,
config=configuration,
drop_rate=configuration.dropout)
elif configuration.model == "LearnGAPoseNet":
posenet = LearnGAPoseNet(feature_extractor=feature_extractor,
drop_rate=configuration.dropout)
dataloader = get_dataloader(configuration)
evaluator = Evaluator(config=configuration,
model=posenet,
def get_args():
parser = argparse.ArgumentParser(description='Training script for mapnet')
parser.add_argument('--config', type=str, default=None)
args = parser.parse_args()
return args
if __name__ == "__main__":
args = get_args()
print("Parse configuration...")
configuration = get_configuration(MapNetConfigurator(), args)
# Model
print("Load model...")
feature_extractor = models.resnet34(pretrained=True)
posenet = PoseNet(feature_extractor, drop_rate=configuration.dropout)
model = MapNet(mapnet=posenet)
param_list = [{'params': model.parameters()}]
kwargs = dict(loss_fn=nn.L1Loss(),
beta=configuration.beta,
gamma=configuration.gamma,
rel_beta=configuration.beta,
rel_gamma=configuration.gamma,
learn_beta=configuration.learn_beta,
learn_rel_beta=configuration.learn_beta)