FlowDiffuser<T>

Models & Types · AiDotNet.Video.Motion

FlowDiffuser diffusion-based optical flow with iterative refinement.

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

For Beginners

FlowDiffuser uses a diffusion process to iteratively refine optical flow estimates. Starting from random noise, it progressively denoises to produce accurate dense flow fields.

How It Works

References:

  • Paper: “FlowDiffuser: Advancing Optical Flow Estimation with Diffusion Models” (Luo et al., CVPR 2024)

FlowDiffuser applies diffusion models to optical flow estimation with iterative refinement, producing accurate and smooth flow fields.

Example

// Create a FlowDiffuser model for diffusion-based optical flow estimation
var flowDiffuser = new FlowDiffuser<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 FlowDiffuser<double>(architecture);

Constructors

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

Methods

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