Tingle Li

I am an incoming Ph.D. student at Berkeley Artificial Intelligence Research (BAIR) Lab, affiliated with UC Berkeley. Currently, I am working with Prof. Hang Zhao from IIIS, Tsinghua University, and Prof. Andrew Owens from The University of Michigan.

Previously, I received my B.E. from Tiangong University, and was a research intern at Duke University (China campus), advised by Prof. Ming Li.

I study the perceptual insights brought by naturally paired data (sight and sound). These may include audio signal processing and audio-visual learning.

     

profile photo
News

2021

2020

Publications
lft Learning Visual Styles from Audio-Visual Associations
Tingle Li, Yichen Liu, Andrew Owens, Hang Zhao
In submission to ECCV, 2022  
PDF / Project Page

We learn from unlabeled data to manipulate the style of an image using sound.

umt Modality Laziness: Everybody's Business is Nobody's Business
Chenzhuang Du, Jiaye Teng, Tingle Li, Yichen Liu, Yue Wang, Yang Yuan, Hang Zhao
In submission to ICML, 2022  
PDF

With multi-modal data as inputs, the encoders from naive fusion training will suffer from learning insufficient representations of each modality.

avd Neural Dubber: Dubbing for Videos According to Scripts
Chenxu Hu, Qiao Tian, Tingle Li, Yuping Wang, Yuxuan Wang, Hang Zhao
NeurIPS, 2021  
PDF / Project Page / Press

Automatic video dubbing driven by a neural network.

CVC CVC: Contrastive Learning for Non-parallel Voice Conversion
Tingle Li, Yichen Liu, Chenxu Hu, Hang Zhao
Interspeech, 2021 (ISCA Student Travel Grant) 
PDF / Project Page / Code

One-way GAN training for non-parallel voice conversion.

sams Sams-Net: A Sliced Attention-based Neural Network for Music Source Separation
Tingle Li, Jiawei Chen, Haowen Hou, Ming Li
ISCSLP, 2021 (Oral, Best Undergraduate Dissertation) 
PDF / Project Page

The scope of attention is narrowed down to the intra-chunk musical features that are most likely to affect each other.

fsc The DKU Speech Activity Detection and Speaker Identification Systems for Fearless Steps Challenge Phase-02
Qingjian Lin, Tingle Li, Ming Li
Interspeech, 2020  
PDF / Leaderboard

SoTA performance for speech activity detection and speaker identification.

atss Atss-Net: Target Speaker Separation via Attention-based Neural Network
Tingle Li, Qingjian Lin, Yuanyuan Bao, Ming Li
Interspeech, 2020  
PDF / Project Page

Adapted Transformer to the speech separation for more efficient and generalizable performance.

oml Optimal Mapping Loss: A Faster Loss for End-to-End Speaker Diarization
Qingjian Lin, Tingle Li, Lin Yang, Junjie Wang, Ming Li
Odyssey, 2020  
PDF

A new mapping loss based on Hungarian algorithm that reduces time complexity while maintaining performance for speaker diarization.


Last updated Mar. 2022.
Template from Jon Barron.