lambda_noobj * loss_noobj + loss_class) / batch_size
total_loss += loss
# backward
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (i + 1) % 5 == 0:
print(
'Epoch [%d/%d], Iter [%d/%d] Loss: %.4f, average_loss: %.4f' %
(epoch + 1, max_epoch, i + 1, len(train_dloader), loss,
total_loss / (i + 1)))
num_iter += 1
model.eval()
with torch.no_grad():
valid_loss = 0
test_accuracy = 0
test_num_data = 0
for x, y in test_dloader:
# implement testing pipeline here
x = x.to(device)
y = y.to(device)
logit = model(x)
loss_xy, loss_wh, loss_obj, loss_noobj, loss_class = compute_loss(
logit, y)
loss = (lambda_coord * (loss_xy + loss_wh) + loss_obj +
lambda_noobj * loss_noobj + loss_class) / batch_size
valid_loss += loss
device = torch.device(
'cuda' if args.use_gpu and torch.cuda.is_available() else 'cpu')
FloatTensor = torch.cuda.FloatTensor if args.use_gpu and torch.cuda.is_available(
) else torch.FloatTensor
classes = utils.load_classes(
args.class_path) # Extracts class labels from file
# Set up model
model = Yolo().to(device)
if args.weights_path is not None:
# Load darknet weights
model.load_darknet_weights(args.weights_path)
model.eval() # Set in evaluation mode
# dataloader = DataLoader(
# ImageFolder(args.image_folder, img_size=args.img_size),
# batch_size=args.batch_size,
# shuffle=False,
# num_workers=args.n_cpu,
# )
if not os.path.exists(args.output_path):
os.makedirs(args.output_path)
if not os.path.exists(args.image_folder):
print('No file or directory with the name {}'.format(
args.image_folder))
exit()
