config.json

{
    "name": "DB",
    "data_loader": {
        "type": "ImageDataset",
        "args": {
            "dataset": {
                "train_data_path": [
                    [
                        "manga109_train.txt"
                    ]
                ],
                "train_data_ratio": [
                    1.0
                ],
                "val_data_path": "./manga109_val.txt", 
                "input_size": 640,
                "img_channel": 3,
                "shrink_ratio": 0.4
            },
            "loader": {
                "validation_split": 0.1,
                "train_batch_size": 10,
                "shuffle": true,
                "pin_memory": false,
                "num_workers": 8
            }
        }
    },
    "arch": {
        "type": "DBModel",
        "args": {
            "backbone": "resnet18",
            "fpem_repeat": 2,
            "pretrained": false,
            "segmentation_head": "FPN",
            "afterburner": "UNET"
        }
    },
    "loss": {
        "type": "DBLoss",
        "args": {
            "alpha": 1.0,
            "beta": 10.0,
            "ohem_ratio": 3
        }
    },
    "optimizer": {
        "type": "Adam",
        "args": {
            "lr": 0.001,
            "weight_decay": 0,
            "amsgrad": true
        }
    },
    "lr_scheduler": {
        "type": "CosineAnnealingLR",
        "args": {
            "T_max": 500
        }
    },
    "trainer": {
        "seed": 2,
        "gpus": [
           0 
        ],
        "epochs": 500,
        "display_interval": 10,
        "show_images_interval": 100,
        "resume_checkpoint": "",
        "finetune_checkpoint": "",
        "output_dir": "output",
        "tensorboard": false,
		"wandb": true,
		"wandb_project": "dbnet",
		"wandb_run_name": "ResNet18_UNET_adam_1e-3_CosineAnnealingLR",
        "metrics": "loss"
    }
}