SpiralConvLayer<T>
Layers · AiDotNet.NeuralNetworks.Layers
Implements spiral convolution for mesh vertex processing.
Inheritance: LayerBase<T> → SpiralConvLayer<T>
Implements: IShapeContract
For Beginners
Unlike image convolutions where neighbors are in a grid, mesh vertices have irregular connectivity. Spiral convolution solves this by:
- Starting at each vertex
- Visiting neighbors in a consistent spiral pattern (like a clock hand)
- Gathering features from each neighbor in order
- Applying learned weights to the ordered features
This creates a consistent “template” for convolution regardless of mesh topology.
Applications:
- 3D shape analysis and classification
- Facial expression recognition
- Body pose estimation
- Medical surface analysis
How It Works
SpiralConvLayer operates on mesh vertices by aggregating features from neighbors in a consistent spiral ordering. This enables translation-equivariant convolutions on irregular mesh structures by defining a canonical ordering of vertex neighbors.
Reference: “Neural 3D Morphable Models: Spiral Convolutional Networks” by Bouritsas et al. Reference: “SpiralNet++: A Fast and Highly Efficient Mesh Convolution Operator” by Gong et al.
Constructors
| Constructor | Summary |
|---|---|
SpiralConvLayer(int inputChannels, int outputChannels, int spiralLength, IActivationFunction<T> activationFunction) |
Eager constructor: allocates and initializes the weight/bias tensors immediately for a known input channel count. |
SpiralConvLayer(int outputChannels, int spiralLength, IActivationFunction<T> activationFunction) |
Initializes a new instance of the SpiralConvLayer class with vector activation. |
SpiralConvLayer(int outputChannels, int spiralLength, IVectorActivationFunction<T> vectorActivation) |
Initializes a new instance of the SpiralConvLayer class with vector activation. |
Properties
| Property | Summary |
|---|---|
bool SupportsTraining { get; } |
Gets a value indicating whether this layer supports training (backpropagation). |
int InputChannels { get; set; } |
Gets the number of input feature channels per vertex. |
int OutputChannels { get; set; } |
Gets the number of output feature channels per vertex. |
int SpiralLength { get; set; } |
Gets the spiral sequence length (number of neighbors in the spiral). |
Methods
| Method | Summary |
|---|---|
Dictionary<string, string> GetParameterRoles() |
Returns parameter roles for per-role optimizer configuration (e.g., weight decay exemption for biases). |
IReadOnlyList<OutputAxisContract> OutputAxesFor(int inputRank) |
|
IReadOnlyList<Tensor<T>> GetTrainableParameters() |
Returns all trainable parameter tensors marked with [TrainableParameter]. |
Tensor<T> GetBiases() |
Gets the bias tensor. |
Tensor<T> GetWeights() |
Gets the weight tensor. |
void ResetState() |
Resets cached state from forward/backward passes. |
void SetSpiralIndices(int[] spiralIndices) |
Sets the spiral indices for the mesh being processed. |
void SetTrainableParameters(IReadOnlyList<Tensor<T>> parameters) |
Replaces trainable parameter tensors (e.g., with ParameterBuffer views). |
void UpdateParameters(T learningRate) |
Updates layer parameters using computed gradients. |
void ZeroGrad() |
Clears all gradient fields discovered by convention ({paramName}Gradient). |