“CN-Celeb”版本间的差异

2019年11月7日 (四) 09:20的版本

Introduction

• CN-Celeb, a large-scale Chinese celebrities dataset published by Center for Speech and Language Technology (CSLT) at Tsinghua University.

Members

• Current：Dong Wang, Yunqi Cai, Lantian Li, Yue Fan, Jiawen Kang
• History：Ziya Zhou, Kaicheng Li, Haolin Chen, Sitong Cheng, Pengyuan Zhang

Description

• Collect audio data of 1,000 Chinese celebrities.
• Automatically clip videos through a pipeline including face detection, face recognition, speaker validation and speaker diarization.
• Create a benchmark database for speaker recognition community.

Basic Methods

• Environments: Tensorflow, PyTorch, Keras, MxNet
• Face detection and tracking: RetinaFace and ArcFace models.
• Active speaker verification: SyncNet model.
• Speaker diarization: UIS-RNN model.
• Double check by speaker recognition: VGG model.
• Input: pictures and videos of POIs (Persons of Interest).
• Output: well-labelled videos of POIs (Persons of Interest).

Reports

Stage report v1.0

Publications

Source Code

• Collection Pipeline: celebrity-audio-collection
• Baseline Systems: kaldi-cn-celeb

Download

• Link

Future Plans

• Augment the database to 10,000 people.
• Build a model between SyncNet and Speaker_Diarization based on LSTM, which can learn the relationship of them.

License

• All the resources contained in the database are free for research institutes and individuals.
• No commerical usage is permitted.

References

• Deng et al., "RetinaFace: Single-stage Dense Face Localisation in the Wild", 2019. [1]
• Deng et al., "ArcFace: Additive Angular Margin Loss for Deep Face Recognition", 2018, [2]
• Wang et al., "CosFace: Large Margin Cosine Loss for Deep Face Recognition", 2018, [3]
• Liu et al., "SphereFace: Deep Hypersphere Embedding for Face Recognition", 2017[4]
• Zhong et al., "GhostVLAD for set-based face recognition", 2018. [5]
• Chung et al., "Out of time: automated lip sync in the wild", 2016.[6]
• Xie et al., "Utterance-level Aggregation For Speaker Recognition In The Wild", 2019. [7]
• Zhang1 et al., "Fully Supervised Speaker Diarization", 2018. [8]