Home

Awesome

Better, Faster, Stronger UIS-RNN

This repository implements some useful features on top of the original UIS-RNN repository. Some of them are described in the following paper: Supervised Online Diarization with Sample Mean Loss for Multi-Domain Data. Here is a list:

Here is a diagram of the Sample Mean Loss:

<p align="center"> <img src="./resources/SML_diag.png" width="450"> </p>

The UIS-RNN was originally proposed in Fully Supervised Speaker Diarization.

<p align="center"> <img src="https://raw.githubusercontent.com/google/uis-rnn/master/resources/uisrnn.gif" width="500"> </p>

Run the demo

To get started, simply run this command:

python3 demo.py --train_iteration=1000 -l=0.001

This will train a UIS-RNN model using data/toy_training_data.npz, then store the model on disk, perform inference on data/toy_testing_data.npz, print the inference results, and save the averaged accuracy in a text file.

P.S.: The files under data/ are manually generated toy data, for demonstration purpose only. These data are very simple, so we are supposed to get 100% accuracy on the testing data.

Arguments

All the other arguments are the same as per the original repository

Citations

Our paper is cited as:

@article{fini2019supervised,
  title={Supervised online diarization with sample mean loss for multi-domain data},
  author={Fini, Enrico and Brutti, Alessio},
  journal={arXiv preprint arXiv:1911.01266},
  year={2019}
}