DiffusionPreprocessorBase<T>
Base Classes · AiDotNet.Diffusion.Preprocessing
Base class for diffusion model condition preprocessors that convert input images into control signals (edge maps, depth maps, pose skeletons, etc.).
Implements: IDataTransformer<T, Tensor<T>, Tensor<T>>
Known derived types: CannyEdgePreprocessor<T>, ColorPalettePreprocessor<T>, ContentShufflePreprocessor<T>, DWPosePreprocessor<T>, DepthEstimationPreprocessor<T>, InpaintingMaskPreprocessor<T>, LineArtPreprocessor<T>, MLSDPreprocessor<T>, MediaPipeFacePreprocessor<T>, NormalMapPreprocessor<T>, OpenPosePreprocessor<T>, QRCodePreprocessor<T>, SAMPreprocessor<T>, ScribblePreprocessor<T>, SemanticSegPreprocessor<T>, ShufflePreprocessor<T>, SoftEdgePreprocessor<T>, TilePreprocessor<T>
For Beginners
These preprocessors are the “preparation step” before using ControlNet. They extract specific features from your image:
- Edge detection: finds outlines and boundaries
- Depth estimation: estimates how far each pixel is
- Pose detection: finds body keypoints
- Segmentation: identifies object regions
The output becomes the “blueprint” that guides image generation.
How It Works
Diffusion preprocessors transform input images into condition maps that guide controlled generation. For example, a Canny edge preprocessor converts a photo into an edge map that ControlNet uses to preserve structure.
Properties
| Property | Summary |
|---|---|
ControlType OutputControlType { get; } |
Gets the control type this preprocessor produces. |
bool IsFitted { get; } |
|
bool SupportsInverseTransform { get; } |
|
int OutputChannels { get; } |
Gets the number of output channels for the control signal. |
int[] ColumnIndices { get; } |
Methods
| Method | Summary |
|---|---|
Tensor<T> FitTransform(Tensor<T> data) |
|
Tensor<T> InverseTransform(Tensor<T> data) |
|
Tensor<T> Transform(Tensor<T> data) |
|
string[] GetFeatureNamesOut(string[] inputFeatureNames) |
|
void Fit(Tensor<T> data) |