config.json

{
  "inputs": [
    "images"
  ],
  "modules": {
    "avg_pool": {
      "config": {
        "args": {
          "output_size": [
            null,
            1
          ]
        }
      },
      "type": "DeepTextRecognition.AdaptiveAvgPoolModule"
    },
    "feature_extraction": {
      "config": {
        "args": {
          "input_channel": 1,
          "output_channel": 512,
          "variant": "DTRB"
        }
      },
      "type": "DeepTextRecognition.ResNetModel"
    },
    "permute": {
      "config": {
        "args": {
          "dims": [
            0,
            3,
            1,
            2
          ]
        }
      },
      "type": "DeepTextRecognition.PermuteModule"
    },
    "prediction": {
      "config": {
        "args": {
          "hidden_size": 256,
          "input_size": 256,
          "num_classes": 96
        }
      },
      "type": "DeepTextRecognition.TextRecognitionAttentionModel"
    },
    "processing": {
      "config": {
        "args": {
          "channels_size": 1,
          "image_size": [
            32,
            100
          ],
          "padding": "left"
        }
      },
      "type": "DeepTextRecognition.ImageProcessor"
    },
    "sequence_modeling": {
      "config": {
        "args": {
          "hidden_sizes": [
            256
          ],
          "input_size": 512,
          "output_size": 256
        }
      },
      "type": "DeepTextRecognition.BiLSTMModel"
    },
    "squeeze": {
      "config": {
        "args": {
          "dim": 3
        }
      },
      "type": "DeepTextRecognition.SqueezeModule"
    },
    "tokenizer": {
      "config": {
        "args": {
          "characters": "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~",
          "max_length": 25
        }
      },
      "type": "DeepTextRecognition.AttentionTokenizer"
    },
    "transformation": {
      "config": {
        "args": {
          "F": 20,
          "I_channel_num": 1,
          "I_r_size": [
            32,
            100
          ],
          "I_size": [
            32,
            100
          ]
        }
      },
      "type": "DeepTextRecognition.TPSModel"
    }
  },
  "order": [
    "processing",
    "transformation",
    "feature_extraction",
    "permute",
    "avg_pool",
    "squeeze",
    "sequence_modeling",
    "prediction",
    "tokenizer"
  ],
  "outputs": [
    "tokenizer:labels"
  ],
  "routing": {
    "avg_pool": {
      "inputs": [
        "permute:permuted_features"
      ],
      "outputs": [
        "avg_pool:pooled_features"
      ]
    },
    "feature_extraction": {
      "inputs": [
        "transformation:transformed_images"
      ],
      "outputs": [
        "feature_extraction:extracted_features"
      ]
    },
    "permute": {
      "inputs": [
        "feature_extraction:extracted_features"
      ],
      "outputs": [
        "permute:permuted_features"
      ]
    },
    "prediction": {
      "inputs": [
        "sequence_modeling:modeled_features"
      ],
      "outputs": [
        "prediction:predictions"
      ]
    },
    "processing": {
      "inputs": [
        "images"
      ],
      "outputs": [
        "processing:processed_images"
      ]
    },
    "sequence_modeling": {
      "inputs": [
        "squeeze:squeezed_features"
      ],
      "outputs": [
        "sequence_modeling:modeled_features"
      ]
    },
    "squeeze": {
      "inputs": [
        "avg_pool:pooled_features"
      ],
      "outputs": [
        "squeeze:squeezed_features"
      ]
    },
    "tokenizer": {
      "inputs": [
        "prediction:predictions"
      ],
      "outputs": [
        "tokenizer:labels"
      ]
    },
    "transformation": {
      "inputs": [
        "processing:processed_images"
      ],
      "outputs": [
        "transformation:transformed_images"
      ]
    }
  }
}