MoMaskModel<T>

Models & Types · AiDotNet.Diffusion.MotionGeneration

MoMask model for masked generative modeling of 3D human motion sequences.

Inheritance: DiffusionModelBase<T> → LatentDiffusionModelBase<T> → MoMaskModel<T>

For Beginners

MoMask generates human motion faster than diffusion-based methods by converting motion into tokens (like words in a sentence) and predicting masked tokens in parallel. This is similar to how BERT fills in missing words, but for body movement.

How It Works

MoMask generates human motion using masked token prediction in a discrete motion token space. It first quantizes motion into tokens via RVQ (residual vector quantization), then uses masked prediction for fast parallel generation.

Reference: Guo et al., “MoMask: Generative Masked Modeling of 3D Human Motions”, CVPR 2024

Example

var options = new DiffusionModelOptions<float> { LatentChannels = 263, DefaultInferenceSteps = 10 };
var noise = Tensor<float>.CreateRandom(new[] { 1, 263, 196 });
var trainX = Tensor<float>.CreateRandom(4, 8);
var trainY = Tensor<float>.CreateRandom(4, 2);
var result = new AiModelBuilder<float, Tensor<float>, Tensor<float>>()
    .ConfigureModel(new MoMaskModel<float>(options: options))
    .Build(trainX, trainY);
var motion = result.Predict(noise);

Constructors

Constructor Summary
MoMaskModel(NeuralNetworkArchitecture<T> architecture, DiffusionModelOptions<T> options, INoiseScheduler<T> scheduler, SiTPredictor<T> predictor, StandardVAE<T> vae, IConditioningModule<T> conditioner, int? seed)

Properties

Property Summary
IConditioningModule<T> Conditioner { get; }
INoisePredictor<T> NoisePredictor { get; }
IVAEModel<T> VAE { get; }
int LatentChannels { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()