VoxCeleb2: Deep Speaker Recognition

2018/06/27 by Joon Son Chung, Arsha Nagrani, Andrew Zisserman · 12 citations
Computer Science · Engineering · #Artificial intelligence #Benchmark (surveying) #Computer science #Convolutional neural network #Feature extraction #Key (lock) #Machine learning #Margin (machine learning) #Music and Audio Processing #Pattern recognition (psychology) #Pipeline (software) #Speaker diarisation #Speaker recognition #Speech Recognition and Synthesis #Speech and Audio Processing #Speech recognition #cs.CV #cs.SD #eess.AS

paper · pdf · doi:10.21437/interspeech.2018-1929

To appear in Interspeech 2018. The audio-visual dataset can be downloaded from http://www.robots.ox.ac.uk/~vgg/data/voxceleb2 . 1806.05622v2: minor fixes; 5 pages

arxiv created 2018/06/27 · openalex publication_date 2018/08/28 · arxiv updated 2020/11/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

<p>The objective of this paper is speaker recognition under noisy and unconstrained conditions.</p> <br/> <p>We make two key contributions. First, we introduce a very large-scale audio-visual speaker recognition dataset collected from open-source media. Using a fully automated pipeline, we curate VoxCeleb2 which contains over a million utterances from over 6,000 speakers. This is several times larger than any publicly available speaker recognition dataset.</p> <br/> <p>Second, we develop and compare Convolutional Neural Network (CNN) models and training strategies that can effectively recognise identities from voice under various conditions. The models trained on the VoxCeleb2 dataset surpass the performance of previous works on a benchmark dataset by a significant margin.</p>

Citations

• Return of the Devil in the Details: Delving Deep into Convolutional Nets
• DeepFace: Closing the Gap to Human-Level Performance in Face Verification
• In Defense of the Triplet Loss for Person Re-Identification
• Deep Speaker: an End-to-End Neural Speaker Embedding System
• MS-Celeb-1M: A Dataset and Benchmark for Large-Scale Face Recognition
• MatConvNet - Convolutional Neural Networks for MATLAB

Cited by

• TidyVoice 2026 Challenge Evaluation Plan
• TidyVoice: A Curated Multilingual Dataset for Speaker Verification Derived from Common Voice
• Disentangle Identity, Cooperate Emotion: Correlation-Aware Emotional Talking Portrait Generation
• MERaLiON-GR: Speech Gender Recognition Model for English and SEA Languages
• Speaker Attribution with Voice Profiles by Graph-Based Semi-Supervised Learning
• Self-Supervised Learning of Interpretable Keypoints From Unlabelled Videos
• Learning spectro-temporal representations of complex sounds with parameterized neural networks
• The VoicePrivacy 2020 Challenge Evaluation Plan
• FineNet: Frame Interpolation and Enhancement for Face Video Deblurring
• Speaker Fuzzy Fingerprints: Benchmarking Text-Based Identification in Multiparty Dialogues
• UniCon: Unified Context Network for Robust Active Speaker Detection
• Bootstrap Equilibrium and Probabilistic Speaker Representation Learning for Self-Supervised Speaker Verification
• Beyond Residual Connections: Manifold-Constrained Hyper-Connections for Robust Speaker Representation Learning