Always convert to RGB

dan/encoder.py

@@ -92,9 +92,7 @@ class FCN_Encoder(Module):
 
         self.init_blocks = ModuleList(
             [
-                ConvBlock(
-                    params["input_channels"], 16, stride=(1, 1), dropout=self.dropout
-                ),
+                ConvBlock(3, 16, stride=(1, 1), dropout=self.dropout),
                 ConvBlock(16, 32, stride=(2, 2), dropout=self.dropout),
                 ConvBlock(32, 64, stride=(2, 2), dropout=self.dropout),
                 ConvBlock(64, 128, stride=(2, 2), dropout=self.dropout),

dan/ocr/document/train.py

@@ -138,7 +138,6 @@ def get_config():
             },
             "transfered_charset": True,  # Transfer learning of the decision layer based on charset of the line HTR model
             "additional_tokens": 1,  # for decision layer = [<eot>, ], only for transferred charset
-            "input_channels": 3,  # number of channels of input image
             "dropout": 0.5,  # dropout rate for encoder
             "enc_dim": 256,  # dimension of extracted features
             "nb_layers": 5,  # encoder

docs/get_started/training.md

@@ -51,7 +51,6 @@ version: 0.0.1
 parameters:
   max_char_prediction: int
   encoder:
-    input_channels: int
     dropout: float
   decoder:
     enc_dim: int

docs/usage/train/parameters.md

@@ -123,7 +123,6 @@ For a detailed description of all augmentation transforms, see the [dedicated pa
 | `model_params.transfer_learning.decoder`  | Model to load for the decoder [state_dict_name, checkpoint_path, learnable, strict]. | `list`        | `["encoder", "pretrained_models/dan_rimes_page.pt", True, False]` |
 | `model_params.transfered_charset`         | Transfer learning of the decision layer based on charset of the model to transfer.   | `bool`        | `True`                                                            |
 | `model_params.additional_tokens`          | For decision layer = [<eot>, ], only for transferred charset.                        | `int`         | `1`                                                               |
-| `model_params.input_channels`             | Number of channels of input image.                                                   | `int`         | `3`                                                               |
 | `model_params.dropout`                    | Dropout probability in the encoder.                                                  | `float`       | `0.5`                                                             |
 | `model_params.enc_dim`                    | Dimension of features extracted by the encoder.                                      | `int`         | `256`                                                             |
 | `model_params.nb_layers`                  | Number of layers in the encoder.                                                     | `int`         | `5`                                                               |

tests/conftest.py

@@ -86,7 +86,6 @@ def training_config():
             "transfer_learning": None,
             "transfered_charset": True,  # Transfer learning of the decision layer based on charset of the line HTR model
             "additional_tokens": 1,  # for decision layer = [<eot>, ], only for transferred charset
-            "input_channels": 3,  # number of channels of input image
             "dropout": 0.5,  # dropout rate for encoder
             "enc_dim": 256,  # dimension of extracted features
             "nb_layers": 5,  # encoder

tests/data/prediction/parameters.yml

@@ -3,7 +3,6 @@ version: 0.0.1
 parameters:
   max_char_prediction: 200
   encoder:
-    input_channels: 3
     dropout: 0.5
   decoder:
     enc_dim: 256