DNN embeddings extraction from audio and speech recordings using PyTorch.

Convert kaldi feature extraction and nnet3 models into Tensorflow Lite models. Currently aimed at converting kaldi's x-vector models and diarization pipelines to tensorflow models.

PyTorch implementation of a Time Delay Neural Network (TDNN)

Implementation of the paper "Spoken Language Recognition using X-vectors" in Pytorch

PyTorch implementation of the Factorized TDNN (TDNN-F) from "Semi-Orthogonal Low-Rank Matrix Factorization for Deep Neural Networks" and Kaldi

Time delay neural network (TDNN) implementation in Pytorch using unfold method