ch_PP-OCRv3_det_cml.yml 4.4 KB
Newer Older
L
LDOUBLEV 已提交
1 2 3 4 5 6
Global:
  debug: false
  use_gpu: true
  epoch_num: 500
  log_smooth_window: 20
  print_batch_step: 10
L
LDOUBLEV 已提交
7
  save_model_dir: ./output/ch_PP-OCR_v3_det/
L
LDOUBLEV 已提交
8 9 10 11 12 13 14 15 16 17 18 19
  save_epoch_step: 100
  eval_batch_step:
  - 0
  - 400
  cal_metric_during_train: false
  pretrained_model: null
  checkpoints: null
  save_inference_dir: null
  use_visualdl: false
  infer_img: doc/imgs_en/img_10.jpg
  save_res_path: ./checkpoints/det_db/predicts_db.txt
  distributed: true
L
LDOUBLEV 已提交
20

L
LDOUBLEV 已提交
21 22 23 24 25 26 27 28 29 30 31 32 33 34
Architecture:
  model_type: det
  algorithm: DB
NetWorks:
  Student:
    model_type: det
    algorithm: DB
    Transform: null
    Backbone:
      name: MobileNetV3
      scale: 0.5
      model_name: large
      disable_se: true
    Neck:
L
fix det  
LDOUBLEV 已提交
35
      name: CAPAN
L
LDOUBLEV 已提交
36
      out_channels: 96
L
fix det  
LDOUBLEV 已提交
37
      shortcut: True
L
LDOUBLEV 已提交
38 39 40 41 42 43 44 45 46 47 48 49 50
    Head:
      name: DBHead
      k: 50
  Student2:
    model_type: det
    algorithm: DB
    Transform: null
    Backbone:
      name: MobileNetV3
      scale: 0.5
      model_name: large
      disable_se: true
    Neck:
L
fix det  
LDOUBLEV 已提交
51
      name: CAPAN
L
LDOUBLEV 已提交
52
      out_channels: 96
L
fix det  
LDOUBLEV 已提交
53
      shortcut: True
L
LDOUBLEV 已提交
54 55 56 57
    Head:
      name: DBHead
      k: 50
  Teacher:
L
LDOUBLEV 已提交
58
    pretrained: ./inference/ch_ppocr_resnet50/student
L
LDOUBLEV 已提交
59 60 61 62 63 64 65
    model_type: det
    algorithm: DB
    Backbone:
      name: ResNet
      in_channels: 3
      layers: 50
    Neck:
L
fix det  
LDOUBLEV 已提交
66
      name: FEPAN
L
LDOUBLEV 已提交
67 68 69
      out_channels: 256
    Head:
      name: DBHead
L
fix det  
LDOUBLEV 已提交
70
      kernel_list: [7,2,2]
L
LDOUBLEV 已提交
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
      k: 50
    
Loss:
  name: CombinedLoss
  loss_config_list:
  - DistillationDilaDBLoss:
      weight: 1.0
      model_name_pairs:
      - ["Student", "Teacher"]
      - ["Student2", "Teacher"]
      key: maps
      balance_loss: true
      main_loss_type: DiceLoss
      alpha: 5
      beta: 10
      ohem_ratio: 3
  - DistillationDMLLoss:
      model_name_pairs:
      - ["Student", "Student2"]
      maps_name: "thrink_maps"
      weight: 1.0
      # act: None
      model_name_pairs: ["Student", "Student2"]
      key: maps
  - DistillationDBLoss:
      weight: 1.0
      model_name_list: ["Student", "Student2"]
      # key: maps
      # name: DBLoss
      balance_loss: true
      main_loss_type: DiceLoss
      alpha: 5
      beta: 10
      ohem_ratio: 3

Optimizer:
  name: Adam
  beta1: 0.9
  beta2: 0.999
  lr:
    name: Cosine
    learning_rate: 0.001
    warmup_epoch: 2
  regularizer:
    name: L2
    factor: 5.0e-05
PostProcess:
  name: DBPostProcess
  thresh: 0.3
  box_thresh: 0.6
  max_candidates: 1000
  unclip_ratio: 1.5
Metric:
  name: DetMetric
  main_indicator: hmean
Train:
  dataset:
    name: SimpleDataSet
L
LDOUBLEV 已提交
129
    data_dir: ./train_data/icdar2015/text_localization/
L
LDOUBLEV 已提交
130
    label_file_list:
L
LDOUBLEV 已提交
131 132
      - ./train_data/icdar2015/text_localization/train_icdar2015_label.txt
    ratio_list: [1.0]
L
LDOUBLEV 已提交
133 134 135 136 137
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - DetLabelEncode: null
L
LDOUBLEV 已提交
138
    - CopyPaste:
L
LDOUBLEV 已提交
139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193
    - IaaAugment:
        augmenter_args:
        - type: Fliplr
          args:
            p: 0.5
        - type: Affine
          args:
            rotate:
            - -10
            - 10
        - type: Resize
          args:
            size:
            - 0.5
            - 3
    - EastRandomCropData:
        size:
        - 960
        - 960
        max_tries: 50
        keep_ratio: true
    - MakeBorderMap:
        shrink_ratio: 0.4
        thresh_min: 0.3
        thresh_max: 0.7
    - MakeShrinkMap:
        shrink_ratio: 0.4
        min_text_size: 8
    - NormalizeImage:
        scale: 1./255.
        mean:
        - 0.485
        - 0.456
        - 0.406
        std:
        - 0.229
        - 0.224
        - 0.225
        order: hwc
    - ToCHWImage: null
    - KeepKeys:
        keep_keys:
        - image
        - threshold_map
        - threshold_mask
        - shrink_map
        - shrink_mask
  loader:
    shuffle: true
    drop_last: false
    batch_size_per_card: 8
    num_workers: 4
Eval:
  dataset:
    name: SimpleDataSet
L
LDOUBLEV 已提交
194
    data_dir: ./train_data/icdar2015/text_localization/
L
LDOUBLEV 已提交
195
    label_file_list:
L
LDOUBLEV 已提交
196
      - ./train_data/icdar2015/text_localization/test_icdar2015_label.txt
L
LDOUBLEV 已提交
197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - DetLabelEncode: null
    - DetResizeForTest: null
    - NormalizeImage:
        scale: 1./255.
        mean:
        - 0.485
        - 0.456
        - 0.406
        std:
        - 0.229
        - 0.224
        - 0.225
        order: hwc
    - ToCHWImage: null
    - KeepKeys:
        keep_keys:
        - image
        - shape
        - polys
        - ignore_tags
  loader:
    shuffle: false
    drop_last: false
    batch_size_per_card: 1
    num_workers: 2