VoxSRC 2020: The Second VoxCeleb Speaker Recognition Challenge

TL;DR

The VoxSRC 2020 challenge evaluated the state-of-the-art in speaker recognition and diarisation using unconstrained YouTube data, providing datasets, benchmarks, and insights into progress since the previous challenge.

Contribution

This paper introduces the second VoxCeleb Speaker Recognition Challenge, including new datasets, evaluation protocols, and a comparison of current methods in unconstrained speaker recognition.

Findings

Significant improvements over the first challenge in speaker recognition accuracy

Effective baseline systems established for diarisation and recognition tasks

Insights into current challenges and future directions in speaker recognition

Abstract

We held the second installment of the VoxCeleb Speaker Recognition Challenge in conjunction with Interspeech 2020. The goal of this challenge was to assess how well current speaker recognition technology is able to diarise and recognize speakers in unconstrained or `in the wild' data. It consisted of: (i) a publicly available speaker recognition and diarisation dataset from YouTube videos together with ground truth annotation and standardised evaluation software; and (ii) a virtual public challenge and workshop held at Interspeech 2020. This paper outlines the challenge, and describes the baselines, methods used, and results. We conclude with a discussion of the progress over the first installment of the challenge.