UPRNet<T>
Models & Types · AiDotNet.Video.FrameInterpolation
UPR-Net: Unified Pyramid Recurrent Network for video frame interpolation.
Inheritance: NeuralNetworkBase<T> → VideoNeuralNetworkBase<T> → FrameInterpolationBase<T> → UPRNet<T>
How It Works
This native graph follows the authors’ released upr_base.py: one shared
three-stage feature pyramid, one shared radius-four partial-correlation motion
estimator, and one shared synthesis U-Net are recurrently reused from coarse to
fine. Motion and synthesis use normalized forward soft splatting. UPR-Net does
not contain a ConvLSTM and does not create separate modules per pyramid level.
Constructors
| Constructor | Summary |
|---|---|
UPRNet(NeuralNetworkArchitecture<T> architecture, UPRNetOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer) |
Creates the paper-faithful native UPR-Net. |
UPRNet(NeuralNetworkArchitecture<T> architecture, string modelPath, UPRNetOptions options) |
Creates the paper-faithful native UPR-Net. |
Methods
| Method | Summary |
|---|---|
ModelMetadata<T> GetModelMetadata() |
|
Tensor<T> Forward(Tensor<T> input) |
Runs the standard midpoint graph. |
Tensor<T> ForwardForTraining(Tensor<T> input) |
|
Tensor<T> Interpolate(Tensor<T> frame0, Tensor<T> frame1, double t) |