dnCNNLayers

Get denoising convolutional neural network layers

Syntax

layers = dnCNNLayers

layers = dnCNNLayers(NetworkDepth=networkDepth)

Description

layers = dnCNNLayers returns layers of the denoising convolutional neural network (DnCNN) for grayscale images.

This function requires Deep Learning Toolbox™.

layers = dnCNNLayers(NetworkDepth=networkDepth) returns a denoising convolutional neural network with networkDepth number of convolutional layers.

Before R2021a, use the equivalent syntax layers = dnCNNLayers("NetworkDepth",networkDepth).

Examples

Get Layers of Image Denoising Network

Get layers of the image denoising convolutional neural network, 'DnCNN'. Request the default number of layers, which returns 20 convolution layers.

layers = dnCNNLayers

ans = 
  1×58 Layer array with layers:

     1   'InputLayer'   Image Input           50×50×1 images
     2   'Conv1'        2-D Convolution       64 3×3×1 convolutions with stride [1  1] and padding [1  1  1  1]
     3   'ReLU1'        ReLU                  ReLU
     4   'Conv2'        2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
     5   'BNorm2'       Batch Normalization   Batch normalization with 64 channels
     6   'ReLU2'        ReLU                  ReLU
     7   'Conv3'        2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
     8   'BNorm3'       Batch Normalization   Batch normalization with 64 channels
     9   'ReLU3'        ReLU                  ReLU
    10   'Conv4'        2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    11   'BNorm4'       Batch Normalization   Batch normalization with 64 channels
    12   'ReLU4'        ReLU                  ReLU
    13   'Conv5'        2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    14   'BNorm5'       Batch Normalization   Batch normalization with 64 channels
    15   'ReLU5'        ReLU                  ReLU
    16   'Conv6'        2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    17   'BNorm6'       Batch Normalization   Batch normalization with 64 channels
    18   'ReLU6'        ReLU                  ReLU
    19   'Conv7'        2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    20   'BNorm7'       Batch Normalization   Batch normalization with 64 channels
    21   'ReLU7'        ReLU                  ReLU
    22   'Conv8'        2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    23   'BNorm8'       Batch Normalization   Batch normalization with 64 channels
    24   'ReLU8'        ReLU                  ReLU
    25   'Conv9'        2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    26   'BNorm9'       Batch Normalization   Batch normalization with 64 channels
    27   'ReLU9'        ReLU                  ReLU
    28   'Conv10'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    29   'BNorm10'      Batch Normalization   Batch normalization with 64 channels
    30   'ReLU10'       ReLU                  ReLU
    31   'Conv11'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    32   'BNorm11'      Batch Normalization   Batch normalization with 64 channels
    33   'ReLU11'       ReLU                  ReLU
    34   'Conv12'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    35   'BNorm12'      Batch Normalization   Batch normalization with 64 channels
    36   'ReLU12'       ReLU                  ReLU
    37   'Conv13'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    38   'BNorm13'      Batch Normalization   Batch normalization with 64 channels
    39   'ReLU13'       ReLU                  ReLU
    40   'Conv14'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    41   'BNorm14'      Batch Normalization   Batch normalization with 64 channels
    42   'ReLU14'       ReLU                  ReLU
    43   'Conv15'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    44   'BNorm15'      Batch Normalization   Batch normalization with 64 channels
    45   'ReLU15'       ReLU                  ReLU
    46   'Conv16'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    47   'BNorm16'      Batch Normalization   Batch normalization with 64 channels
    48   'ReLU16'       ReLU                  ReLU
    49   'Conv17'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    50   'BNorm17'      Batch Normalization   Batch normalization with 64 channels
    51   'ReLU17'       ReLU                  ReLU
    52   'Conv18'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    53   'BNorm18'      Batch Normalization   Batch normalization with 64 channels
    54   'ReLU18'       ReLU                  ReLU
    55   'Conv19'       2-D Convolution       64 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]
    56   'BNorm19'      Batch Normalization   Batch normalization with 64 channels
    57   'ReLU19'       ReLU                  ReLU
    58   'Conv20'       2-D Convolution       1 3×3×64 convolutions with stride [1  1] and padding [1  1  1  1]

Input Arguments

networkDepth — Number of convolution layers

20 (default) | positive integer

Number of convolution layers, specified as a positive integer with value greater than or equal to 3.

Data Types: single | double | int8 | int16 | int32 | uint8 | uint16 | uint32

Output Arguments

layers — Network layers

vector of Layer objects

Denoising convolutional neural network layers, returned as a vector of Layer (Deep Learning Toolbox) objects.

Tips

• Train a custom image denoising network by providing these layers to the trainnet (Deep Learning Toolbox) function (since R2024b). Specify the loss function as "mse" for regression.

  layers = dnCNNLayers;
  net = trainnet(data,layers,"mse",options);

• The DnCNN network can detect noise and other high-frequency image artifacts. For example, you can train the DnCNN network to increase image resolution or remove JPEG compression artifacts. The example JPEG Image Deblocking Using Deep Learning shows how to train a DnCNN to reduce JPEG compression artifacts in an image.

References

[1] Zhang, K., W. Zuo, Y. Chen, D. Meng, and L. Zhang. "Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising." IEEE Transactions on Image Processing. Vol. 26, Issue 7, 2017, pp. 3142–3155.

Version History

Introduced in R2017b

R2024b: Support for training using trainnet

Starting in R2024b, the dnCNNLayers function does not include a final regression layer. The new output of dnCNNLayers now supports training a custom image denoising network using the trainnet (Deep Learning Toolbox) function.

Before, the function returned a vector of layers with a RegressionOutputLayer (Deep Learning Toolbox) as the final layer, and you trained the custom denoising network using the trainNetwork (Deep Learning Toolbox) function. The RegressionOutputLayer object and the trainNetwork function are no longer recommended.

To update your code, replace instances of the trainNetwork function with trainnet, and specify the third argument of trainnet as the loss function "mse".

Prior Usage (Before R2024b)	Replacement Code
This example trains a custom denoising network using the trainNetwork function, where the layers returned by dnCNNLayers includes a regression output layer. layers = dnCNNLayers; trainedNet = trainNetwork(data,layers,options);	This example trains a custom denoising network using the trainnet function, where the layers returned by dnCNNLayers no longer includes a regression output layer. Specify the third argument of trainnet as the loss function "mse". layers = dnCNNLayers; trainedNet = trainnet(data,layers,"mse",options);

If you need to reproduce the prior training workflow using the trainNetwork function, then you can create and concatenate a regression output layer to the layers returned by dnCNNLayers. For example:

layers = dnCNNLayers;
layers = [layers, RegressionOutputLayer(Name="FinalRegressionLayer")];
trainedNet = trainNetwork(data,layers,options);