pytorch_connectomics/tutorials/mito_mitolab.yaml at master · PytorchConnectomics/pytorch_connectomics · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
_base_:
  - ../connectomics/config/all_profiles.yaml

experiment_name: cem-mitolab_mednext2d
description: Mitochondria segmentation on CEM-MitoLab dataset using MedNeXt 2D

default:
  model:
    arch:
      profile: mednext_s
    out_channels: 3
    loss:
      deep_supervision: true
      losses:
      - function: DiceLoss
        weight: 1.0
        kwargs: {include_background: false, sigmoid: true, smooth_nr: 1e-5, smooth_dr: 1e-5}
        pred_slice: "0:1"
        target_slice: "0:1"
      - function: BCEWithLogitsLoss
        weight: 0.5
        pred_slice: "0:1"
        target_slice: "0:1"
      - function: BCEWithLogitsLoss
        weight: 0.5
        pred_slice: "1:2"
        target_slice: "1:2"
      - function: WeightedMSELoss
        weight: 1.0
        kwargs: {tanh: true}
        pred_slice: "2:3"
        target_slice: "2:3"
    mednext:
      size: S
      kernel_size: 3
      dim: 2d
  data:
    image_transform:
      clip_percentile_low: 0.0
      clip_percentile_high: 1.0
    label_transform:
      targets:
      - name: binary
      - name: instance_boundary
        kwargs:
          thickness: 1
          edge_mode: seg-all
          mode: 2d
      - name: instance_edt
        kwargs:
          mode: 2d
          quantize: false
    data_transform:
      pad_size:
      - 0
      - 0
    dataloader:
      use_preloaded_cache_train: false
      use_preloaded_cache_val: false
      patch_size:
      - 224
      - 224
  inference:
    system:
      num_workers: 4
    test_time_augmentation:
      enabled: false
      channel_activations:
        - {channels: "0:2", activation: sigmoid}
        - {channels: "2:3", activation: tanh}
    prediction_transform:
      enabled: true
      intensity_scale: 255
      intensity_dtype: uint8
    save_prediction:
      enabled: true

train:
  system:
    num_gpus: 1
    num_workers: 0
    seed: 42
  optimization:
    max_epochs: 100
    n_steps_per_epoch: -1
    accumulate_grad_batches: 1
    precision: bf16-mixed
    optimizer:
      lr: 0.001
      weight_decay: 0.01
      eps: 1.0e-08
    scheduler:
      name: StepLR
      params:
        step_size: 100000
        gamma: 1.0
  monitor:
    logging:
      scalar:
        loss:
        - train_loss_total_epoch
        loss_every_n_steps: 50
      images:
        max_images: 16
        num_slices: 1
        log_every_n_epochs: 1
    checkpoint:
      monitor: val/loss
      save_top_k: 3
      save_every_n_epochs: 1
      dirpath: outputs/cem-mitolab_mednext2d/checkpoints/
      checkpoint_filename: epoch={epoch:03d}-val_loss={val/loss:.4f}
    early_stopping:
      monitor: in_loss_total_epoch
      patience: 20
      min_delta: 0.0001
      threshold: 0.05
      divergence_threshold: 2.0
  data:
    train:
      dataset_type: filename
      json: datasets/cem-mitolab/files.json
      image_key: images
      label_key: masks
      split_ratio: 0.9
      resolution:
      - 0
      - 5
      - 5

test:
  inference:
    save_prediction:
      output_path: outputs/cem-mitolab_mednext2d/results/
  data:
    test:
      image: datasets/cem-mitolab/test.json
      label:
      resolution:
      - 0
      - 5
      - 5
  evaluation:
    enabled: false
    metrics: []