ICDAR2019RecTS

This is my code about the task1 and task2 of the ICDAR 2019 Robust Reading Challenge on Reading Chinese Text on Signboard

Docs

ICDARRecTs_task2

• ICDARRecTs_2Configure.py:
  • This is the configuration file, but I can't use it to configure in this code.
• ICDARRecTs_2DataSet.py:
  • This the dataset file, the class ICDARRecTs_2DataSet is the subclass of the Dataset
    • img_path(str): refer to the path of the image.
    • dictionary Optional(dict): refer to the dictionary that use to decode the char to number,but it isn't deprecated to use.
    • coordinates_path Optional(str): refer to the path of the coordinate of the textline.
    • img_transform(Callable): refer to the image's transfomation function.
    • train(bool): refer to the model is train or test.
    • load_img(Callable): refer to the load a image function.
• ICDARRecTs_2NN.py:
  • This is the model file, it includes 'DenseLSTM','VGGLSTM','DenseCNN','VGGFC','ResNetLSTM'.
• ICDARRecTs_2Preprocessing.py:
  • This is the preprocessing file, it mainly includes the function of rotate_crop, get_dictionary
    • rotate_crop: it uses to rectify the textline according to the coordinates.
    • get_dictionary: it uses to get the dictionary of the char.
• ICDARRecTs_2Test.py:
  • DEVICE: refer to test on a CPU or GPU device.
  • BATCH_SIZE: refer to batch size.
  • PATH: refer to the path of dictionary.
  • DICTIONARY_NAME: refer to the name of the dictionary.
  • COORDINATES_PATH: refer to the path of the coordinate of the textline.
  • IMAGE_PATH: refer to the path of the image.
  • MODEL_PATH: refer to the path of the parameters of the model.
  • MODEL_NAME: refer to the name of the parameters of the model.
  • NUM_CLASS: refer to the number of the classification.
• ICDARRecTs_2Train.py:
  • DEVICE: refer to test on a CPU or GPU device
  • BATCH_SIZE: refer to the batch size.
  • EPOCH: refer to the loop count。
  • PATH：refer to the path of dictionary.
  • DICTIONARY_NAME：refer to the path of the coordinate of the textline.
  • IMAGE_PATH：refer to the path of the image.
  • MODEL_PATH：refer to the path of the parameters of the model.
  • MODEL_NAME：refer to the name of the parameters of the model.
  • PRETRAIN：refer to whether there is a saved parameters of the model.
  • NUM_CLASS: refer to the number of the classification.
  • LR: refer to the initial the learning rate.
  • MAX_ACCURACY: refer to the current max accuracy.