diff --git a/dan/encoder.py b/dan/encoder.py
index 6057caddb1e96bcb6a6d0c1e2e38cb637e2757a0..0e38f6330b2035b75f1ca4ed87ff446ccb3c0015 100644
--- a/dan/encoder.py
+++ b/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),
diff --git a/dan/ocr/document/train.py b/dan/ocr/document/train.py
index ad4643b92506fe65befca106430d36ea1bcfa146..97cae76166000d011095d5b50d58c38263b272b0 100644
--- a/dan/ocr/document/train.py
+++ b/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
diff --git a/docs/get_started/training.md b/docs/get_started/training.md
index 15106d40c17bd3f5c3639ef04323b9e759a2fe8d..b70992810b035d0a2fbe1b3acc255fb7e60423fe 100644
--- a/docs/get_started/training.md
+++ b/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
diff --git a/docs/usage/train/parameters.md b/docs/usage/train/parameters.md
index ac4026fe205029a84d239d98ac63aa3f2de4c8b0..b8dbef3cec08c564dc5fc848d82ee23eb4675605 100644
--- a/docs/usage/train/parameters.md
+++ b/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`                                                               |
diff --git a/tests/conftest.py b/tests/conftest.py
index e804cb36734d4d7eb8af90ea2057b8d14e308ff0..4a365be7b90aefcab64fcf500da664ccea4c0b20 100644
--- a/tests/conftest.py
+++ b/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
diff --git a/tests/data/prediction/parameters.yml b/tests/data/prediction/parameters.yml
index 2bad8803509baf24851dbeb2cf192da1ed6ae800..32ffff56398fee1ac0efd7b129684b24605c8eda 100644
--- a/tests/data/prediction/parameters.yml
+++ b/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