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": {
          "N_max_character": 26,
          "n_class": 38,
          "n_position": 26
        }
      },
      "type": "DeepTextRecognition.SRNDecoder"
    },
    "processing": {
      "config": {
        "args": {
          "channels_size": 1,
          "image_size": [
            32,
            100
          ],
          "padding": "left"
        }
      },
      "type": "DeepTextRecognition.ImageProcessor"
    },
    "sequence_modeling": {
      "config": {
        "args": {
          "d_inner": 1024,
          "d_k": 64,
          "d_model": 512,
          "d_v": 64,
          "dropout": 0.1,
          "n_head": 8,
          "n_layers": 2,
          "n_position": 26
        }
      },
      "type": "DeepTextRecognition.TransformerEncoderv1"
    },
    "squeeze": {
      "config": {
        "args": {
          "dim": 3
        }
      },
      "type": "DeepTextRecognition.SqueezeModule"
    },
    "tokenizer": {
      "config": {
        "args": {
          "characters": [
            "0",
            "1",
            "2",
            "3",
            "4",
            "5",
            "6",
            "7",
            "8",
            "9",
            "a",
            "b",
            "c",
            "d",
            "e",
            "f",
            "g",
            "h",
            "i",
            "j",
            "k",
            "l",
            "m",
            "n",
            "o",
            "p",
            "q",
            "r",
            "s",
            "t",
            "u",
            "v",
            "w",
            "x",
            "y",
            "z"
          ],
          "max_length": 25,
          "pad_token": 36
        }
      },
      "type": "DeepTextRecognition.SRNTokenizer"
    }
  },
  "order": [
    "processing",
    "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": [
        "processing:processed_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_1",
        "prediction:predictions_2",
        "prediction:predictions_3"
      ]
    },
    "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_3"
      ],
      "outputs": [
        "tokenizer:labels"
      ]
    }
  }
}