RRDBNetGenerator<T>
Layers · AiDotNet.NeuralNetworks.Layers
RRDBNet Generator - the full generator architecture from ESRGAN and Real-ESRGAN.
Inheritance: LayerBase<T> → RRDBNetGenerator<T>
Implements: IShapeContract
For Beginners
This is the “brain” of Real-ESRGAN that transforms low-resolution images into high-resolution ones.
Key components:
- Initial Conv: Extracts basic features from the input image
- RRDB Blocks: 23 deep blocks that learn how to enhance details
- Trunk Conv + Residual: Combines deep features with initial features
- Upsampling: Makes the image bigger (2x or 4x depending on scale)
- Final Convs: Produces the final RGB output image
The default parameters (64 features, 32 growth, 23 RRDBs, 4x scale) are from the Real-ESRGAN paper and produce excellent results for general image super-resolution.
How It Works
This implements the complete RRDBNet generator from the ESRGAN paper (Wang et al., 2018). It combines multiple RRDB blocks with upsampling for image super-resolution.
The architecture is:
Reference: Wang et al., “Real-ESRGAN: Training Real-World Blind Super-Resolution
with Pure Synthetic Data”, ICCV 2021. https://arxiv.org/abs/2107.10833
Constructors
| Constructor | Summary |
|---|---|
RRDBNetGenerator(int inputChannels, int outputChannels, int numFeatures, int growthChannels, int numRRDBBlocks, int scale, double residualScale) |
Initializes a new RRDBNet generator. |
Properties
| Property | Summary |
|---|---|
bool SupportsTraining { get; } |
|
int NumFeatures { get; } |
Gets the number of feature channels. |
int NumRRDBBlocks { get; } |
Gets the number of RRDB blocks. |
int Scale { get; } |
Gets the upscaling factor. |
Methods
| Method | Summary |
|---|---|
IReadOnlyList<ILayer<T>> GetSubLayers() |
|
IReadOnlyList<OutputAxisContract> OutputAxesFor(int inputRank) |
|
void ResetState() |
|
void UpdateParameters(T learningRate) |