IVideoVAEModel<T>

Interfaces · AiDotNet.Interfaces

A VAE that encodes video, [batch, channels, frames, height, width], compressing time as well as space.

Implements: IVAEModel<T>

Known implementations: Causal3DVAE<T>, TemporalInterpolationVAE<T>

For Beginners

An image VAE shrinks each picture; a video VAE also merges neighbouring frames, so a clip of 49 frames might become 13 latent frames. GetLatentFrameCount says how many latent frames a clip gives.

Properties

Property Summary
int TemporalCompression { get; } Gets how many input frames one latent frame covers (after the first, for a causal VAE).

Methods

Method Summary
int GetLatentFrameCount(int frames) Gets the number of latent frames a clip of frames frames encodes to.