parser_train = subparsers.add_parser('train') parser_train.add_argument('--port', type=int, default=80) parser_train.add_argument('--root', required=True) parser_train.add_argument('--epochs', type=int, default=100) parser_train.add_argument('--workers', type=int, default=4) parser_train.add_argument('--batch', type=int, default=1) parser_train.add_argument('--steps-loss', type=int, default=50) parser_train.add_argument('--steps-plot', type=int, default=0) parser_train.add_argument('--steps-save', type=int, default=500) return parser.parse_args() NUM_CHANNELS = 3 NUM_CLASSES = 2 color_transform = Colorize() image_transform = ToPILImage() input_transform = Compose([ Scale(512), CenterCrop(512), ToTensor(), Normalize([.485, .456, .406], [.229, .224, .225]), ]) target_transform = Compose([ Scale(512), CenterCrop(512), ToLabel(), Relabel(255, 1) ])
from torch.utils.data import DataLoader from torchvision.transforms import Compose, Normalize from torchvision.transforms import ToTensor, ToPILImage, Resize sys.path.append("/home/nico/PycharmProjects/project-marvel/defect-detection") from defect_detection.evaluator.evaluation import save_metrics_on_results from piwise.criterion import CrossEntropyLoss2d from piwise.dataset import VOCTrain, VOCTest from piwise.network import FCN8, FCN16, FCN32, UNet, PSPNet, SegNet from piwise.transform import ToLabel, Colorize from piwise.visualize import Dashboard NUM_CHANNELS = 3 NUM_CLASSES = 16 color_transform = Colorize(n=NUM_CLASSES) image_transform = ToPILImage() input_transform = Compose([ Resize(256), ToTensor(), # Normalize([.485, .456, .406], [.229, .224, .225]), Normalize([.5, .5, .5], [.5, .5, .5]), ]) target_transform = Compose([ Resize(256), ToLabel(), # Relabel(255, 21), ]) def train(args, model):