SKFlow<T>

Models & Types · AiDotNet.Video.Motion

SKFlow selective kernel attention for efficient optical flow estimation.

Inheritance: NeuralNetworkBase<T> → VideoNeuralNetworkBase<T> → OpticalFlowBase<T> → SKFlow<T>

For Beginners

SKFlow uses selective kernels that adapt their receptive field size based on local motion patterns. Different kernel sizes handle different motion scales automatically.

How It Works

References:

  • Paper: “SKFlow: Learning Optical Flow with Super Kernels” (Sun et al., 2022)

SKFlow uses selective kernel attention mechanisms to efficiently capture multi-scale motion information for optical flow estimation.

Example

// Create an SKFlow model for selective-kernel optical flow estimation
var skFlow = new SKFlow<double>();
            
// Or configure with custom parameters
var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.ThreeDimensional,
    taskType: NeuralNetworkTaskType.Regression,
    inputHeight: 256, inputWidth: 256, inputDepth: 3, outputSize: 2);
var model = new SKFlow<double>(architecture);

Constructors

Constructor Summary
SKFlow() Creates a new SKFlow model for native training and inference.
SKFlow(NeuralNetworkArchitecture<T> architecture, SKFlowOptions options)

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> EstimateFlow(Tensor<T> frame0, Tensor<T> frame1)