McnetOptions
Options & Configuration · AiDotNet.Video.Options
Configuration for Mcnet.
Inheritance: ModelOptions → NeuralNetworkOptions → McnetOptions
For Beginners
These control how deep each of the two pathways is, how many past frames the model watches, how many future frames it predicts, and how much the critic network influences training.
How It Works
Defaults follow Villegas, Yang, Hong, Lin and Lee, “Decomposing Motion and Content for Natural Video Sequence Prediction” (ICLR 2017, arXiv:1706.08033): learning rate 1e-4, alpha = 1, beta = 0.02 (KTH) or 0.001 (UCF-101), lambda = 1, p = 2, and 10 input frames on KTH / 4 on UCF-101.
This replaces DRVIOptions. Its NumContentBlocks and NumMotionBlocks survive in
spirit because MCnet genuinely has two encoder pathways, but the model they configured was an
interpolator, not a predictor.
Constructors
| Constructor | Summary |
|---|---|
McnetOptions() |
Properties
| Property | Summary |
|---|---|
OnnxModelOptions OnnxOptions { get; set; } |
Gets or sets ONNX runtime options. |
double AdversarialLossWeight { get; set; } |
Gets or sets beta, the adversarial weight. |
double DropoutRate { get; set; } |
Gets or sets the dropout rate. |
double GradientLossExponent { get; set; } |
Gets or sets lambda, the gradient-difference exponent. |
double ImageLossWeight { get; set; } |
Gets or sets alpha, the image-loss weight. |
double LearningRate { get; set; } |
Gets or sets the learning rate. |
int NumContentBlocks { get; set; } |
Gets or sets the depth of the content pathway (f_cont). |
int NumDecoderBlocks { get; set; } |
Gets or sets the decoder depth (g_dec). |
int NumFeatures { get; set; } |
Gets or sets the feature width of the encoders and decoder. |
int NumInputFrames { get; set; } |
Gets or sets how many past frames the model observes. |
int NumMotionBlocks { get; set; } |
Gets or sets the depth of the motion pathway (f_dyn). |
int NumPredictedFrames { get; set; } |
Gets or sets how many future frames to predict. |
int NumScales { get; set; } |
Gets or sets the number of scales at which residual features are communicated (r_t^l). |
int PixelLossNorm { get; set; } |
Gets or sets p, the pixel-loss norm. |
string ModelPath { get; set; } |
Gets or sets an optional ONNX model path for inference-only use. |