PaddleOCR训练自己的文字识别模型（数码管AI读数）

PaddlePaddle建议版本：2.0.2

优质博客：

alt alt alt 上述训练数据结构安排行不通的话，如下设置：

配置文件： alt

train.txt中的内容： alt

alt alt alt alt

alt

alt alt

  ...
  # Add a custom dictionary, such as modify the dictionary, please point the path to the new dictionary
  character_dict_path: ppocr/utils/ppocr_keys_v1.txt
  # Modify character type
  character_type: ch
  ...
  # Whether to recognize spaces
  use_space_char: True


Optimizer:
  ...
  # Add learning rate decay strategy
  lr:
    name: Cosine
    learning_rate: 0.001
  ...

...

Train:
  dataset:
    # Type of dataset，we support LMDBDataSet and SimpleDataSet
    name: SimpleDataSet
    # Path of dataset
    data_dir: ./train_data/
    # Path of train list
    label_file_list: ["./train_data/train_list.txt"]
    transforms:
      ...
      - RecResizeImg:
          # Modify image_shape to fit long text
          image_shape: [3, 32, 320]
      ...
  loader:
    ...
    # Train batch_size for Single card
    batch_size_per_card: 256
    ...

Eval:
  dataset:
    # Type of dataset，we support LMDBDataSet and SimpleDataSet
    name: SimpleDataSet
    # Path of dataset
    data_dir: ./train_data
    # Path of eval list
    label_file_list: ["./train_data/val_list.txt"]
    transforms:
      ...
      - RecResizeImg:
          # Modify image_shape to fit long text
          image_shape: [3, 32, 320]
      ...
  loader:
    # Eval batch_size for Single card
    batch_size_per_card: 256
    ...

注意预测/评估的配置文件必须与训练一致。

3 评价

可以通过修改文件中的Eval.dataset.label_file_list字段来设置评估数据集configs/rec/rec_icdar15_train.yml。

python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec_icdar15_train.yml -o Global.checkpoints={path/to/weights}/best_accuracy

4 预测

4.1 训练引擎预测

使用paddleocr训练的模型，可以通过以下脚本快速得到预测。

默认预测图片存储在中infer_img，权重通过指定-o Global.checkpoints：

python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.pretrained_model={path/to/weights}/best_accuracy Global.load_static_weights=false Global.infer_img=doc/imgs_words/en/word_1.jpg

获取输入图像的预测结果：

infer_img: doc/imgs_words/en/word_1.png
        result: ('joint', 0.9998967)

用于预测的配置文件必须与训练一致。比如你用完成了中文模型的训练python3 tools/train.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml，可以使用如下命令来预测中文模型：

python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.pretrained_model={path/to/weights}/best_accuracy Global.load_static_weights=false Global.infer_img=doc/imgs_words/ch/word_1.jpg

github链接