configs/rec/sar/resnet31_lstm_sar.yml

Global:
  device: gpu
  epoch_num: 20
  log_smooth_window: 20
  print_batch_step: 10
  output_dir: ./output/rec/u14m_filter/resnet31_lstm_sar
  eval_epoch_step: [0, 1]
  eval_batch_step: [0, 500]
  cal_metric_during_train: True
  pretrained_model:
  checkpoints:
  use_tensorboard: false
  infer_img:
  # for data or label process
  character_dict_path: ./tools/utils/EN_symbol_dict.txt
  max_text_length: 25
  use_space_char: False
  save_res_path: ./output/rec/u14m_filter/predicts_resnet31_lstm_sar.txt
  use_amp: True
  grad_clip_val: 1.0

Optimizer:
  name: Adam
  lr: 0.002 # for 4gpus bs256/gpu
  weight_decay: 0.0
  filter_bias_and_bn: False

LRScheduler:
  name: OneCycleLR
  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
  cycle_momentum: False

Architecture:
  model_type: rec
  algorithm: SAR
  Transform:
  Encoder:
    name: ResNet31
  Decoder:
    name: SARDecoder
    mask: True
    use_lstm: True

Loss:
  name: ARLoss

PostProcess:
  name: ARLabelDecode

Metric:
  name: RecMetric
  main_indicator: acc
  is_filter: True

Train:
  dataset:
    name: LMDBDataSet
    data_dir: ../Union14M-L-LMDB-Filtered
    transforms:
      - DecodeImage: # load image
          img_mode: BGR
          channel_first: False
      - PARSeqAug:
      - ARLabelEncode: # Class handling label
      - RobustScannerRecResizeImg:
          image_shape: [3, 48, 48, 160] # h:48 w:[48,160]
          width_downsample_ratio: 0.25
      - KeepKeys:
          keep_keys: ['image', 'label', 'length', 'valid_ratio'] # dataloader will return list in this order
  loader:
    shuffle: True
    batch_size_per_card: 256
    drop_last: True
    num_workers: 4

Eval:
  dataset:
    name: LMDBDataSet
    data_dir: ../evaluation
    transforms:
      - DecodeImage: # load image
          img_mode: BGR
          channel_first: False
      - ARLabelEncode: # Class handling label
      - RobustScannerRecResizeImg:
          image_shape: [3, 48, 48, 160]
          width_downsample_ratio: 0.25
      - KeepKeys:
          keep_keys: ['image', 'label', 'length', 'valid_ratio'] # dataloader will return list in this order
  loader:
    shuffle: False
    drop_last: False
    batch_size_per_card: 256
    num_workers: 2