deepof.models

deep autoencoder models for unsupervised pose detection.

• VQ-VAE: a variational autoencoder with a vector quantization latent-space (https://arxiv.org/abs/1711.00937).

• VaDE: a variational autoencoder with a Gaussian mixture latent-space.

• Contrastive: an embedding model consisting of a single encoder, trained using a contrastive loss.

Functions

get_TCN_decoder(input_shape, latent_dim[, ...])	Return a Temporal Convolutional Network (TCN) decoder.
get_TCN_encoder(input_shape, ...[, use_gnn, ...])	Return a Temporal Convolutional Network (TCN) encoder.
get_recurrent_decoder(input_shape, latent_dim)	Return a recurrent neural decoder.
get_recurrent_encoder(input_shape, ...[, ...])	Return a deep recurrent neural encoder.
get_transformer_decoder(input_shape, latent_dim)	Build a Transformer decoder.
get_transformer_encoder(input_shape, ...[, ...])	Build a Transformer encoder.
get_vade(input_shape, edge_feature_shape, ...)	Build a Gaussian mixture variational autoencoder (VaDE) model, adapted to the DeepOF setting.
get_vqvae(input_shape, edge_feature_shape, ...)	Build a Vector-Quantization variational autoencoder (VQ-VAE) model, adapted to the DeepOF setting.

Classes

Classifier(*args, **kwargs)	Classifier for supervised pose motif elucidation.
Contrastive(*args, **kwargs)	Self-supervised contrastive embeddings.
GaussianMixtureLatent(*args, **kwargs)	Gaussian Mixture probabilistic latent space model.
VQVAE(*args, **kwargs)	VQ-VAE model adapted to the DeepOF setting.
VaDE(*args, **kwargs)	Gaussian Mixture Variational Autoencoder for pose motif elucidation.
VectorQuantizer(*args, **kwargs)	Vector quantizer layer.