STMFNetOptions
Options & Configuration · AiDotNet.Video.Options
Configuration options for STMFNet spatio-temporal multi-flow network.
Inheritance: ModelOptions → NeuralNetworkOptions → STMFNetOptions
For Beginners
Instead of guessing one “best” motion for each pixel, STMFNet makes multiple guesses (flows) and then picks the best one for each part of the image. This works much better in tricky areas like where objects overlap or where motion is ambiguous.
How It Works
STMFNet (2022) uses multiple optical flows in spatio-temporal space:
- Multi-flow estimation: estimates multiple (typically 4) optical flow fields between the
input frames, each capturing different motion hypotheses for ambiguous regions like occlusion boundaries, transparent objects, and repeating textures
- Spatio-temporal feature volume: constructs a 4D (height x width x time x channel) feature
volume from the input frames and all estimated flow fields, capturing the full motion context in a unified representation
- Flow selection network: a learned network that selects the best flow hypothesis for each
pixel by comparing warped features from each flow field, choosing the one that produces the most consistent result
- Residual refinement: after flow-based warping, a refinement network corrects remaining
artifacts using the multi-flow feature volume as context
Constructors
| Constructor | Summary |
|---|---|
STMFNetOptions() |
Initializes a new instance with default values. |
STMFNetOptions(STMFNetOptions 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 LearningRate { get; set; } |
Gets or sets the learning rate. |
int NumFeatures { get; set; } |
Gets or sets the number of feature channels. |
int NumFlowHypotheses { get; set; } |
Gets or sets the number of flow hypotheses per pixel. |
int NumFusionBlocks { get; set; } |
Gets or sets the number of spatio-temporal fusion blocks. |
int NumPyramidLevels { get; set; } |
Gets or sets the number of pyramid levels for multi-scale flow. |
int NumRefineBlocks { get; set; } |
Gets or sets the number of refinement blocks. |
string ModelPath { get; set; } |
Gets or sets the path to the ONNX model file. |