print('Model loaded from {}'.format(args.resume))
else:
print("=> no checkpoint found at '{}'".format(args.resume))
else:
start_epoch = 0
start_step = 0
train_image_paths, train_mask_paths = get_images(image_dir, args.preprocess, phase='train', healthy_included=args.healthyincluded)
train_image_paths2, train_mask_paths2, train_predicted_mask_paths2 = get_images_diaretAL(image_dir2, predicted_dir, args.preprocess, phase='train')
eval_image_paths, eval_mask_paths = get_images(image_dir, args.preprocess, phase='eval', healthy_included=args.healthyincluded)
if net_name == 'unet':
# not using unet, deprecated yet.
train_dataset = IDRIDDataset(train_image_paths, train_mask_paths, 4, transform=
Compose([
RandomRotation(rotation_angle),
RandomCrop(image_size),
]))
eval_dataset = IDRIDDataset(eval_image_paths, eval_mask_paths, 4, transform=
Compose([
RandomCrop(image_size),
]))
elif net_name == 'hednet':
train_dataset1 = IDRIDDataset(train_image_paths, train_mask_paths, 4, transform=
Compose([
RandomRotation(rotation_angle),
RandomCrop(image_size),
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
]))
train_dataset2 = DiaretALDataset(train_image_paths2, train_mask_paths2, train_predicted_mask_paths2, 4, transform=
Compose([
checkpoint = torch.load(args.model)
model.load_state_dict(checkpoint['state_dict'])
print('Model loaded from {}'.format(args.model))
else:
print("=> no checkpoint found at '{}'".format(args.model))
sys.exit(0)
eval_image_paths, eval_mask_paths = get_images(
image_dir,
args.preprocess,
phase='eval',
healthy_included=args.healthyincluded)
if net_name == 'unet':
eval_dataset = IDRIDDataset(eval_image_paths,
eval_mask_paths,
4,
transform=Compose([]))
elif net_name == 'hednet':
eval_dataset = IDRIDDataset(eval_image_paths,
eval_mask_paths,
4,
transform=Compose([
Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
]))
eval_loader = DataLoader(eval_dataset, args.batchsize, shuffle=False)
dice_coeffs_soft, dice_coeffs_hard, vis_images = eval_model(
model, eval_loader)
print(dice_coeffs_soft, dice_coeffs_hard)
#logger.image_summary('eval_images', vis_images, step=0)
checkpoint = torch.load(resume)
start_epoch = checkpoint['epoch']+1
start_step = checkpoint['step']
model.load_state_dict(checkpoint['state_dict'])
g_optimizer.load_state_dict(checkpoint['optimizer'])
print('Model loaded from {}'.format(resume))
else:
print("=> no checkpoint found at '{}'".format(resume))
else:
start_epoch = 0
start_step = 0
train_image_paths, train_mask_paths = get_images(image_dir, args.preprocess, phase='train')
eval_image_paths, eval_mask_paths = get_images(image_dir, args.preprocess, phase='eval')
train_dataset = IDRIDDataset(train_image_paths, train_mask_paths, config.LESION_IDS[args.lesion], transform=
Compose([
RandomRotation(rotation_angle),
RandomCrop(image_size),
]))
eval_dataset = IDRIDDataset(eval_image_paths, eval_mask_paths, config.LESION_IDS[args.lesion])
train_loader = DataLoader(train_dataset, batchsize, shuffle=True)
eval_loader = DataLoader(eval_dataset, batchsize, shuffle=False)
g_scheduler = lr_scheduler.StepLR(g_optimizer, step_size=200, gamma=0.9)
criterion = nn.CrossEntropyLoss(weight=torch.FloatTensor(config.CROSSENTROPY_WEIGHTS).to(device))
train_model(model, args.lesion, args.preprocess, train_loader, eval_loader, criterion, g_optimizer, g_scheduler, \
batchsize, num_epochs=config.EPOCHES, start_epoch=start_epoch, start_step=start_step)
start_epoch = 0
start_step = 0
train_image_paths, train_mask_paths = get_images(image_dir,
config.PREPROCESS,
phase='train')
eval_image_paths, eval_mask_paths = get_images(image_dir,
config.PREPROCESS,
phase='eval')
if net_name == 'unet':
train_dataset = IDRIDDataset(
train_image_paths,
train_mask_paths,
config.CLASS_ID,
transform=Compose([
RandomRotation(rotation_angle),
#ColorJitter(brightness=0.2, contrast=0.2, saturation=0.2, hue=0.2),
RandomCrop(image_size),
]))
eval_dataset = IDRIDDataset(eval_image_paths,
eval_mask_paths,
config.CLASS_ID,
transform=Compose([
RandomCrop(image_size),
]))
elif net_name == 'hednet':
train_dataset = IDRIDDataset(
train_image_paths,
train_mask_paths,
config.CLASS_ID,
model = HNNNet(pretrained=True, class_number=2)
resume = args.model
if os.path.isfile(resume):
print("=> loading checkpoint '{}'".format(resume))
checkpoint = torch.load(resume)
start_epoch = checkpoint['epoch'] + 1
start_step = checkpoint['step']
try:
model.load_state_dict(checkpoint['state_dict'])
except:
model.load_state_dict(checkpoint['g_state_dict'])
print('Model loaded from {}'.format(resume))
else:
print("=> no checkpoint found at '{}'".format(resume))
model.to(device)
test_image_paths, test_mask_paths = get_images(image_dir,
config.PREPROCESS,
phase='test')
test_dataset = IDRIDDataset(test_image_paths, test_mask_paths, config.LESION_IDS[args.lesion], \
transform=Compose([Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),]))
test_loader = DataLoader(test_dataset, 1, shuffle=False)
auc_result = eval_model(model, test_loader)
print(auc_result)
np.random.seed(args.seed)
torch.manual_seed(args.seed)
if torch.cuda.is_available():
torch.cuda.manual_seed(args.seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
model = UNet(n_channels=3, n_classes=2)
resume = args.model
if os.path.isfile(resume):
print("=> loading checkpoint '{}'".format(resume))
checkpoint = torch.load(resume)
model.load_state_dict(checkpoint['state_dict'])
print('Model loaded from {}'.format(resume))
else:
print("=> no checkpoint found at '{}'".format(resume))
model.to(device)
test_image_paths, test_mask_paths = get_images(image_dir,
args.preprocess,
phase='test')
test_dataset = IDRIDDataset(test_image_paths, test_mask_paths,
config.LESION_IDS[args.lesion])
test_loader = DataLoader(test_dataset, 1, shuffle=False)
auc_result = eval_model(model, test_loader)
print(auc_result)