MoGOptions

Options & Configuration · AiDotNet.Video.Options

Configuration options for MoG motion-aware generative frame interpolation.

Inheritance: ModelOptions → NeuralNetworkOptions → MoGOptions

For Beginners

MoG combines two approaches: first it figures out how things move (optical flow), then uses a generative AI model (diffusion) to “paint” the intermediate frame guided by that motion information. This produces sharper results than just blending warped frames, especially for complex motions.

How It Works

MoG (2025) combines flow estimation with diffusion-based generation:

  • Motion-aware conditioning: first estimates bidirectional optical flow using an EMA-VFI-style

flow network, then uses the estimated flows as spatial conditioning for a diffusion model rather than directly warping frames

  • Flow-conditioned diffusion: the denoising U-Net receives concatenated flow maps as

additional input channels, guiding the diffusion process to generate motion-consistent intermediate frames with fine texture details

  • Generative refinement: instead of blending warped frames (which can produce ghosting),

the diffusion model generates the intermediate frame from scratch, conditioned on the input frames and estimated motion, producing sharp results even in occluded regions

  • Progressive denoising: multi-step denoising with motion-aware noise scheduling that

preserves motion coherence in early steps and refines textures in later steps

Constructors

Constructor Summary
MoGOptions() Initializes a new instance with default values.
MoGOptions(MoGOptions other) Initializes a new instance by copying from another instance.

Properties

Property Summary
OnnxModelOptions OnnxOptions { get; set; } Gets or sets the ONNX runtime options.
VideoModelVariant Variant { get; set; } Gets or sets the model variant.
double DropoutRate { get; set; } Gets or sets the dropout rate.
double GuidanceScale { get; set; } Gets or sets the guidance scale for classifier-free guidance.
double LearningRate { get; set; } Gets or sets the learning rate.
int NumDiffusionSteps { get; set; } Gets or sets the number of diffusion denoising steps during inference.
int NumFeatures { get; set; } Gets or sets the number of feature channels.
int NumFlowScales { get; set; } Gets or sets the number of flow estimation scales.
int NumResBlocks { get; set; } Gets or sets the number of U-Net residual blocks per level.
string ModelPath { get; set; } Gets or sets the path to the ONNX model file.