import torch, os, imageio, re, sys
import numpy as np
import torchvision as tv
import torchvision.transforms as T
from PIL import Image
sys.path.append('..')
import transforms as tr
#ROTATE_512 = tr.Compose([
# tr.RandomRotate(),
# tr.Lift(T.ToTensor()),
# tr.RandomCropTransform((512, 512))
#])
#CROP_512 = tr.Compose([
# tr.Lift(T.ToTensor()),
# tr.RandomCropTransform((512, 512)),
#])
UNCUT = tr.Compose([
# tr.Lift(T.Pad(88)),
tr.Lift(T.ToTensor())
])
weights[background] = n * self.bg_weight / background.sum().item()
weights[foreground] = n * self.fg_weight / foreground.sum().item()
mask = mask * weights
img = torch.cat([img, self.mean[None]], dim=0)
if self.flip:
img, mask, lbl = random_flip(img, mask, lbl)
return img, mask, lbl
ROTATE_TRANS_1024 = tr.Compose([
tr.AspectPreservingResizeTransform((1024, 768)),
tr.Lift(T.Pad(88)),
tr.RandomRotate(),
])
PAD_TRANS_1024 = tr.Compose([
tr.AspectPreservingResizeTransform((1024, 768)),
tr.Lift(T.Pad(88)),
])
RotatedISICDataset = rotated_dataset(ISICDataset)
if __name__ == '__main__':
target_size = 1024, 768
img_transform = T.Compose([T.ColorJitter(0.3, 0.3, 0.3, 0.), T.ToTensor()])
import sys
sys.path.append('..')
from utils import rotated_dataset
RotatedDeepglobeDataset = rotated_dataset(DeepglobeDataset)
if __name__ == '__main__':
import sys
import matplotlib.pyplot as plt
import torchvision.transforms as T
sys.path.append('..')
import transforms as tr
d = RotatedDeepglobeDataset(
'/home/jatentaki/Storage/jatentaki/Datasets/roads/deepglobe/test/',
global_transform=tr.Lift(T.ToTensor()))
for i in range(25, 50):
img, mask, lbl = d[i]
fig, (a1, a2, a3) = plt.subplots(1, 3)
img = img.numpy().transpose(1, 2, 0)
a1.imshow(img)
a2.imshow(lbl.numpy()[0])
a3.imshow(mask.numpy()[0])
plt.show()
if args.action == 'inspect' and args.batch_size != 1:
args.batch_size = 1
print("Setting --batch-size to 1 for inspection")
if args.action != 'train' and args.epochs is not None:
print("Ignoring --epochs outside of training mode")
if args.no_jit and args.optimize:
print("Ignoring --optimize in --no-jit setting")
writer.add_text('general', str(vars(args)))
transform = T.Compose([T.CenterCrop(644), T.ToTensor()])
test_global_transform = tr.Lift(T.Pad(40))
tr_global_transform = [
# tr.RandomRotate(),
# tr.RandomFlip(),
tr.Lift(T.Pad(40))
]
tr_global_transform = tr.Compose(tr_global_transform)
train_data = loader.DriveDataset(args.data_path,
training=True,
bloat=args.bloat,
from_=args.cut,
img_transform=transform,
mask_transform=transform,
label_transform=transform,
if args.action == 'inspect' and args.batch_size != 1:
args.batch_size = 1
print("Setting --batch-size to 1 for inspection")
if args.action != 'train' and args.epochs is not None:
print("Ignoring --epochs outside of training mode")
if args.no_jit and args.optimize:
print("Ignoring --optimize in --no-jit setting")
writer.add_text('general', str(vars(args)))
transform = T.Compose([T.CenterCrop(644), T.ToTensor()])
# if we are not padding the convolutions, we have to pad the input
aug_pad = None if args.padding else tr.Lift(T.Pad(40))
test_global_transform = aug_pad
tr_global_transform = tr.Compose(
[tr.RandomRotate(), tr.RandomFlip(), aug_pad])
train_data = loader.DriveDataset(args.data_path,
training=True,
bloat=args.bloat,
from_=args.cut,
img_transform=transform,
mask_transform=transform,
label_transform=transform,
global_transform=tr_global_transform)
train_loader = DataLoader(train_data,
cuda = torch.cuda.is_available()
n_params = 0
for param in network.parameters():
n_params += param.numel()
print(n_params, 'learnable parameters')
if cuda:
network = network.cuda()
checkpoint = framework.load_checkpoint(args.load)
_, _, model_dict, _ = checkpoint
network.load_state_dict(model_dict)
tentrans = tr.Compose([tr.Lift(T.Pad(88)), tr.Lift(T.ToTensor())])
#threshold = 0.6#0.6364
with torch.no_grad():
for i, ((path, img), ) in enumerate(tqdm(loader)):
if i == args.early_stop:
break
scaler = APRTrans((1024, 768))
downsized = scaler.forward(img)
input = tentrans(downsized)[0]
input = data.add_mean(input)[None]
if cuda:
input = input.cuda()
prediction = network(input)