e2e_r50_vd_pg.yml

Global:
  use_gpu: False
  epoch_num: 600
  log_smooth_window: 20
  print_batch_step: 2
  save_model_dir: ./output/pg_r50_vd_tt/
  save_epoch_step: 1
  # evaluation is run every 5000 iterationss after the 4000th iteration
  eval_batch_step: [ 0, 1000 ]
  # if pretrained_model is saved in static mode, load_static_weights must set to True
  load_static_weights: False
  cal_metric_during_train: False
  pretrained_model:
  checkpoints:
  save_inference_dir:
  use_visualdl: False
  infer_img:
  save_res_path: ./output/pg_r50_vd_tt/predicts_pg.txt

Architecture:
  model_type: e2e
  algorithm: PG
  Transform:
  Backbone:
    name: ResNet
    layers: 50
  Neck:
    name: PGFPN
    model_name: large
  Head:
    name: PGHead
    model_name: large

Loss:
  name: PGLoss

#Optimizer:
#  name: Adam
#  beta1: 0.9
#  beta2: 0.999
#  lr:
#    name: Cosine
#    learning_rate: 0.001
#    warmup_epoch: 1
#  regularizer:
#    name: 'L2'
#    factor: 0

Optimizer:
  name: RMSProp
  lr:
    name: Piecewise
    learning_rate: 0.001
    decay_epochs: [ 40, 80, 120, 160, 200 ]
    values: [ 0.001, 0.00033, 0.0001, 0.000033, 0.00001 ]
  regularizer:
    name: 'L2'
    factor: 0.00005

PostProcess:
  name: PGPostProcess
  score_thresh: 0.8
  cover_thresh: 0.1
  nms_thresh: 0.2

Metric:
  name: E2EMetric
  main_indicator: f_score_e2e

Train:
  dataset:
    name: PGDateSet
    label_file_list:
    ratio_list:
    data_format: textnet  # textnet/partvgg
    Lexicon_Table: [ '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z' ]
    transforms:
      - DecodeImage: # load image
          img_mode: BGR
          channel_first: False
      - PGProcessTrain:
          batch_size: 14
          data_format: icdar
          tcl_len: 64
          min_crop_size: 24
          min_text_size: 4
          max_text_size: 512
      - KeepKeys:
          keep_keys: [ 'images', 'tcl_maps', 'tcl_label_maps', 'border_maps','direction_maps', 'training_masks', 'label_list', 'pos_list', 'pos_mask' ] # dataloader will return list in this order
  loader:
    shuffle: True
    drop_last: True
    batch_size_per_card: 1
    num_workers: 8

Eval:
  dataset:
    name: PGDateSet
    data_dir: ./train_data/
    label_file_list:
    transforms:
      - DecodeImage: # load image
          img_mode: BGR
          channel_first: False
      - E2ELabelEncode:
          label_list: [ '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z' ]
      - E2EResizeForTest:
          valid_set: totaltext
          max_side_len: 768
      - NormalizeImage:
          scale: 1./255.
          mean: [ 0.485, 0.456, 0.406 ]
          std: [ 0.229, 0.224, 0.225 ]
          order: 'hwc'
      - ToCHWImage:
      - KeepKeys:
          keep_keys: [ 'image', 'shape', 'polys', 'strs', 'tags' ]
  loader:
    shuffle: False
    drop_last: False
    batch_size_per_card: 1 # must be 1
    num_workers: 2