VideoHyperparameterOptions
Options & Configuration · AiDotNet.Video.Options
Shared configuration for video models: action recognition, segmentation, generation, super-resolution, frame interpolation, tracking, denoising, inpainting and depth.
Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → VideoHyperparameterOptions
Known derived types: BasicVSRPlusPlusOptions, CutieOptions, DIFRINTOptions, DKMOptions, DPFlowOptions, DepthAnythingV2Options, E2FGVIOptions, EDVROptions, FILMOptions, FastDVDNetOptions, FlowDiffuserOptions, FlowFormerOptions, FlowFormerPlusPlusOptions, GMFlowOptions, InternVideo2Options, MemFlowOptions, MiDaSOptions, NeuFlowV2Options, ProPainterOptions, RAFTOptions, RAPIDFlowOptions, RIFEOptions, RPKNetOptions, RVMOptions, RealESRGANOptions, RoMaOptions, SAM2Options, SEARAFTOptions, SKFlowOptions, SlowFastOptions, TimeSformerOptions, UFMOptions, UniMatchOptions, VRTOptions, VideoCLIPVideoOptions, VideoFlowOptions, VideoMAEOptions, XMemOptions
For Beginners
Video models work on several frames at once rather than a single picture, so on top of the usual size settings they need to know how many frames they look at together. Each model’s own options class ships the values from its paper, so you do not normally set any of these.
How It Works
Derived options classes assign their paper’s values in their parameterless constructor.
See ModelHyperparameterOptions for why these properties are non-nullable.
Not to be confused with VideoModelOptions<T>, which is an earlier and
now largely abandoned attempt at the same idea: it takes a type parameter it never uses,
follows the nullable + Effective* pattern this design moves away from, and is
derived from by exactly one of the 108 options classes under src/Video/Options
(96 of them derive straight from NeuralNetworkOptions instead). It is left in place
for now and removed when the video models are wired to this base.
Properties
| Property | Summary |
|---|---|
int EmbedDim { get; set; } |
Gets or sets the width of the embedding produced for each token or patch. |
int NumClasses { get; set; } |
Gets or sets the number of output categories, for classification models. |
int NumFeatures { get; set; } |
Gets or sets the width of the model’s main feature representation. |
int NumFrames { get; set; } |
Gets or sets how many consecutive frames the model processes together. |
int NumHeads { get; set; } |
Gets or sets the number of attention heads. |
int NumIterations { get; set; } |
Gets or sets the number of refinement passes, for models that iterate towards an answer (optical flow, diffusion sampling). |
int NumLayers { get; set; } |
Gets or sets the number of stacked blocks in the model. |
int ScaleFactor { get; set; } |
Gets or sets the upscaling factor, for super-resolution models. |