def get_loss_train(model, data_train, criterion, wo_mask=True):
"""
Calculate loss over train set
"""
model.eval()
total_acc = 0
total_loss = 0
for batch, (images, masks) in enumerate(data_train):
with torch.no_grad():
images = Variable(images.cuda())
masks = Variable(masks.cuda())
outputs = model(images)
if wo_mask:
cropped_images = extract_img(outputs.size(2), images)
loss = criterion(outputs, cropped_images)
else:
loss = criterion(outputs, masks)
if outputs.size(1) > 1:
preds = torch.argmax(outputs, dim=1).float()
else:
preds = (outputs > 0).float()
acc = accuracy_check_for_batch(masks.cpu(), preds.cpu(), images.size()[0])
total_acc = total_acc + acc
total_loss = total_loss + loss.cpu().item()
return total_acc/(batch+1), total_loss/(batch + 1)
def validate_model(model, data_val, criterion, epoch, make_prediction=True, save_folder_name='prediction'):
"""
Validation run
"""
# calculating validation loss
total_val_loss = 0
total_val_acc = 0
for batch, (images_v, masks_v) in enumerate(data_val):
with torch.no_grad():
images_v = Variable(images_v.cuda())
masks_v = Variable(masks_v.cuda())
#print(images_v.shape, masks_v.shape)
outputs_v = model(images_v)
total_val_loss = total_val_loss + criterion(outputs_v, masks_v).cpu().item()
#print('out', outputs_v.shape)
outputs_v = torch.argmax(outputs_v, dim=1).float()
#print(outputs_v.shape)
if make_prediction:
im_name = batch # TODO: Change this to real image name so we know
pred_msk = save_prediction_image(outputs_v, im_name, epoch, save_folder_name)
#acc_val = accuracy_check(masks_v, pred_msk)
acc_val = accuracy_check_for_batch(masks_v.cpu(), outputs_v.cpu(), images_v.size()[0])
total_val_acc = total_val_acc + acc_val
return total_val_acc/(batch + 1), total_val_loss/((batch + 1))
def get_loss_train(model, data_train, criterion):
"""
Calculate loss over train set
"""
model.eval() # 测试时加上，不启用 BatchNormalization 和 Dropout
total_acc = 0
total_loss = 0
for batch, (images, masks) in enumerate(data_train):
with torch.no_grad(): #推理阶段（预测阶段），不更新梯度
images = Variable(images.cuda())
masks = Variable(masks.cuda())
outputs = model(images)
loss = criterion(outputs, masks)
preds = torch.argmax(outputs, dim=1).float() # 选概率高的类别
acc = accuracy_check_for_batch(masks.cpu(), preds.cpu(), images.size()[0]) # 计算该batch的acc，images.size()[0]=batch_size
total_acc = total_acc + acc
total_loss = total_loss + loss.cpu().item()
return total_acc/(batch+1), total_loss/(batch + 1)
def get_loss_train(model, data_train, criterion):
"""
Calculate loss over train set
"""
model.eval()
total_acc = 0
total_loss = 0
for batch, (images, masks) in enumerate(data_train):
with torch.no_grad():
images = Variable(images)
masks = Variable(masks)
images.shape
masks.shape
outputs = model(images)
loss = criterion(outputs, masks)
preds = torch.argmax(outputs, dim=1).float()
acc = accuracy_check_for_batch(masks.cpu(), preds.cpu(),
images.size()[0])
total_acc = total_acc + acc
total_loss = total_loss + loss.cpu().item()
return total_acc / (batch + 1), total_loss / (batch + 1)
