model_name: selimsef_spacenet4_resnet34unet model_path: train: false infer: true pretrained: true nn_framework: torch batch_size: 42 data_specs: width: 384 height: 384 dtype: image_type: zscore rescale: false rescale_minima: auto rescale_maxima: auto additional_inputs: channels: 4 label_type: mask is_categorical: false mask_channels: 3 val_holdout_frac: 0.2 data_workers: 12 training_data_csv: '/path/to/training_df.csv' validation_data_csv: inference_data_csv: '/path/to/test_df.csv' training_augmentation: augmentations: RandomScale: scale_limit: - 0.5 - 1.5 interpolation: nearest Rotate: limit: - 5 - 6 border_mode: constant p: 0.3 RandomCrop: height: 416 width: 416 always_apply: true p: 1.0 Normalize: mean: - 0.006479 - 0.009328 - 0.01123 - 0.02082 std: - 0.004986 - 0.004964 - 0.004950 - 0.004878 max_pixel_value: 65535.0 p: 1.0 p: 1.0 shuffle: true validation_augmentation: augmentations: CenterCrop: height: 384 width: 384 p: 1.0 Normalize: mean: - 0.006479 - 0.009328 - 0.01123 - 0.02082 std: - 0.004986 - 0.004964 - 0.004950 - 0.004878 max_pixel_value: 65535.0 p: 1.0 p: 1.0 inference_augmentation: augmentations: Normalize: mean: - 0.006479 - 0.009328 - 0.01123 - 0.02082 std: - 0.004986 - 0.004964 - 0.004950 - 0.004878 max_pixel_value: 65535.0 p: 1.0 p: 1.0 training: epochs: 70 steps_per_epoch: optimizer: AdamW lr: 2e-4 opt_args: weight_decay: 0.0001 loss: focal: dice: loss_weights: focal: 1 dice: 1 metrics: training: validation: checkpoint_frequency: 10 callbacks: lr_schedule: schedule_type: 'arbitrary' schedule_dict: milestones: - 1 - 5 - 15 - 30 - 50 - 60 gamma: 0.5 model_checkpoint: filepath: 'selimsef_resnet34_best.pth' monitor: val_loss model_dest_path: 'selimsef_resnet34.pth' verbose: true inference: window_step_size_x: window_step_size_y: output_dir: 'inference_out/'