PaddleX/zhch/table_structure_recognition/SLANet_plus_paddleocr.yml

Global:
  model_name: SLANet_plus # To use static model for inference.
  use_gpu: False          # 使用CPU
  epoch_num: 8              # 可适当增加训练轮数
  log_smooth_window: 10   # 日志平滑窗口
  print_batch_step: 10    # 每10个batch打印一次日志
  save_model_dir: ./output/SLANet_plus
  save_epoch_step: 2      # 每2个epoch保存一次模型
  # evaluation is run every 331 iterations after the 0th iteration
  eval_batch_step: [0, 50]  # 更频繁评估
  cal_metric_during_train: True
  pretrained_model: "https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/SLANet_plus_pretrained.pdparams"
  checkpoints: 
  save_inference_dir: ./output/SLANet_plus/infer
  use_visualdl: False
  infer_img: paddlex/repo_manager/repos/PaddleOCR/ppstructure/docs/table/table.jpg
  # for data or label process
  character_dict_path: paddlex/repo_manager/repos/PaddleOCR/ppocr/utils/dict/table_structure_dict_ch.txt
  character_type: en
  max_text_length: &max_text_length 500
  box_format: &box_format xyxyxyxy # 'xywh', 'xyxy', 'xyxyxyxy'
  infer_mode: False
  use_sync_bn: False      # 是否使用同步BN, Mac必须关闭
  save_res_path: output/infer
  d2s_train_image_shape: [3, 488, 488]

Optimizer:
  name: Adam
  beta1: 0.9
  beta2: 0.999
  clip_norm: 5.0
  lr:
    learning_rate: 0.0008    # 微调学习率
  regularizer:
    name: 'L2'
    factor: 0.00000

Architecture:
  model_type: table
  algorithm: SLANet
  Backbone:
    name: PPLCNet
    scale: 1.0
    pretrained: True
    use_ssld: True
  Neck:
    name: CSPPAN
    out_channels: 96
  Head:
    name: SLAHead
    hidden_size: 256
    max_text_length: *max_text_length
    loc_reg_num: &loc_reg_num 8

Loss:
  name: SLALoss
  structure_weight: 1.0
  loc_weight: 2.0
  loc_loss: smooth_l1

PostProcess:
  name: TableLabelDecode
  merge_no_span_structure: &merge_no_span_structure True

Metric:
  name: TableMetric
  main_indicator: acc
  compute_bbox_metric: False
  loc_reg_num: *loc_reg_num
  box_format: *box_format
  del_thead_tbody: True

Train:
  dataset:
    name: PubTabDataSet
    data_dir: dataset/table_rec_dataset_examples/          # ✅ 正确路径
    label_file_list: [dataset/table_rec_dataset_examples/train.txt]  # ✅ 正确文件
    transforms:
      - DecodeImage:
          img_mode: BGR
          channel_first: False
      - TableLabelEncode:
          learn_empty_box: False
          merge_no_span_structure: *merge_no_span_structure
          replace_empty_cell_token: False
          loc_reg_num: *loc_reg_num
          max_text_length: *max_text_length
      - TableBoxEncode:
          in_box_format: *box_format
          out_box_format: *box_format
      - ResizeTableImage:
          max_len: 488
      - NormalizeImage:
          scale: 1./255.
          mean: [0.485, 0.456, 0.406]
          std: [0.229, 0.224, 0.225]
          order: 'hwc'
      - PaddingTableImage:
          size: [488, 488]
      - ToCHWImage:
      - KeepKeys:
          keep_keys: ['image', 'structure', 'bboxes', 'bbox_masks', 'length', 'shape']
  loader:
    shuffle: True
    batch_size_per_card: 32            # 减小batch_size提高稳定性
    drop_last: True                    # 丢弃不完整的最后一个batch
    num_workers: 2                     # 减少worker数量，避免CPU负载过高

Eval:
  dataset:
    name: PubTabDataSet
    data_dir: dataset/table_rec_dataset_examples/          # ✅ 正确路径
    label_file_list: [dataset/table_rec_dataset_examples/val.txt]    # ✅ 正确文件
    ratio_list: [1.0]  # 添加这一行，长度与 label_file_list 相同
    transforms:
      - DecodeImage:
          img_mode: BGR
          channel_first: False
      - TableLabelEncode:
          learn_empty_box: False
          merge_no_span_structure: *merge_no_span_structure
          replace_empty_cell_token: False
          loc_reg_num: *loc_reg_num
          max_text_length: *max_text_length
      - TableBoxEncode:
          in_box_format: *box_format
          out_box_format: *box_format
      - ResizeTableImage:
          max_len: 488
      - NormalizeImage:
          scale: 1./255.
          mean: [0.485, 0.456, 0.406]
          std: [0.229, 0.224, 0.225]
          order: 'hwc'
      - PaddingTableImage:
          size: [488, 488]
      - ToCHWImage:
      - KeepKeys:
          keep_keys: ['image', 'structure', 'bboxes', 'bbox_masks', 'length', 'shape']
  loader:
    shuffle: False
    drop_last: False
    batch_size_per_card: 48
    num_workers: 1