parser.add_argument('--no_build_summary', action='store_true', help='Dont save sammary when training to save space')
parser.add_argument('--save_ckpt_freq', default=10, type=int, help='Save checkpoint frequency')
parser.add_argument('--evaluate_only', action='store_true', help='Evaluate pretrained models')
parser.add_argument('--no_validate', action='store_true', help='No validation')
parser.add_argument('--strict', action='store_true', help='Strict mode when loading checkpoints')
parser.add_argument('--val_metric', default='epe', help='Validation metric to select best model')
args = parser.parse_args()
logger = utils.get_logger()
utils.check_path(args.checkpoint_dir)
utils.save_args(args)
filename = 'command_test.txt' if args.mode == 'test' else 'command_train.txt'
utils.save_command(args.checkpoint_dir, filename)
def main():
# For reproducibility
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
np.random.seed(args.seed)
torch.backends.cudnn.benchmark = True
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# Train loader
train_transform_list = [transforms.RandomCrop(args.img_height, args.img_width),
transforms.RandomColor(),
parser.add_argument('--save_type', default='png', choices=['pfm', 'png', 'npy'], help='Save file type')
parser.add_argument('--visualize', action='store_true', help='Visualize disparity map')
# Log
parser.add_argument('--count_time', action='store_true', help='Inference on a subset for time counting only')
parser.add_argument('--num_images', default=100, type=int, help='Number of images for inference')
args = parser.parse_args()
model_name = os.path.basename(args.pretrained_aanet)[:-4]
model_dir = os.path.basename(os.path.dirname(args.pretrained_aanet))
args.output_dir = os.path.join(args.output_dir, model_dir + '-' + model_name)
utils.check_path(args.output_dir)
utils.save_command(args.output_dir)
def main():
# For reproducibility
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
np.random.seed(args.seed)
torch.backends.cudnn.benchmark = True
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# Test loader
test_transform = transforms.Compose([
transforms.ToTensor(),
# 尝试分布式训练
local_master = True if not args.distributed else args.local_rank == 0
utils.save_args(args) if local_master else None
# 打印所用的参数
if local_master:
logger.info('[Info] used parameters: {}'.format(vars(args)))
torch.backends.cudnn.benchmark = True # https://blog.csdn.net/byron123456sfsfsfa/article/details/96003317
utils.check_path(args.checkpoint_dir)
utils.save_args(args) if local_master else None
filename = 'command_test.txt' if args.mode == 'test' else 'command_train.txt'
utils.save_command(args.checkpoint_dir, filename) if local_master else None
def main():
# For reproducibility
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
np.random.seed(args.seed)
train_loader, val_loader = getDataLoader(args, logger)
# Network
aanet = nets.AANet(
args.max_disp,
num_downsample=args.num_downsample,
feature_type=args.feature_type,