NeuFlowV2<T>

Models & Types · AiDotNet.Video.Motion

NeuFlow v2 high-efficiency optical flow on edge devices.

Inheritance: NeuralNetworkBase<T> → VideoNeuralNetworkBase<T> → OpticalFlowBase<T> → NeuFlowV2<T>

For Beginners

NeuFlow V2 is a fast, lightweight optical flow estimator designed for real-time applications. It achieves good accuracy with significantly reduced computation compared to transformer-based methods.

How It Works

References:

  • Paper: “NeuFlow v2: Push High-Efficiency Optical Flow To the Limit” (Zhang et al., 2024)

NeuFlow v2 achieves high-efficiency optical flow estimation suitable for edge devices through a lightweight backbone and optimized inference.

Example

// Create a NeuFlow V2 model for efficient edge-device optical flow
var neuFlow = new NeuFlowV2<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 NeuFlowV2<double>(architecture);

Constructors

Constructor Summary
NeuFlowV2() Creates a new NeuFlowV2 model for native training and inference.
NeuFlowV2(NeuralNetworkArchitecture<T> architecture, NeuFlowV2Options options)

Methods

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