google / uis-rnn

This is the library for the Unbounded Interleaved-State Recurrent Neural Network (UIS-RNN) algorithm, corresponding to the paper Fully Supervised Speaker Diarization.
https://arxiv.org/abs/1810.04719
Apache License 2.0
1.56k stars 319 forks source link

Use a real sequence accuracy evaluation instead of an approximate accuracy #3

Closed wq2012 closed 6 years ago

wq2012 commented 6 years ago

Currently in the demo we use a greedy method to compute sequence match accuracy. It's just an approximation, as an estimation of how well the predicted sequence matches the ground truth.

The correct way should be using Hungarian algorithm to compute the optimal match, which is more meaningful.