Anurag Kumar
Research Lead and Scientist, Meta
anuragkr [AT] ieee [DOT] org
Bio
I am a research lead and scientist at Meta Research. My broad research interests include Deep Learning, Audio/Speech Processing and Multimodal Learning. Often, my research focuses on weakly, self-supervised and unsupervised learning methods for different domains and problems.
Before joining Meta, I finished my PhD from School of Computer Science at Carnegie Mellon University in 2018. I was advised by Prof. Bhiksha Raj. My PhD thesis was Acoustic Intelligence in Machines, and it introduced weakly labeled learning of sounds, which has since then played a crucial role in scaling sound event detection and classification. I obtained my undegraduate degree in Electrical Engineering from Indian Institute of Technology (IIT), Kanpur in 2013.
Some of my recent works have focused on Multimodal Understanding and Generation (audio-only and multimodal) [ Neurips-2023, CVPR-2023, CVPR-2022, IJCAI-2020, ICML-2020]; Speech Enhancement (single chanel, multi-channel, audio-visual) [ICASSP-2023, IEEE JSTSP-2022, ICASSP-2022, ICASSP-2021, ASRU-2021]; Deep Learning based Speech Assessment (Quality and Intelligibility) [ICASSP-2023, Interspeech-2022, Neurips-2021]. Check out my Google Scholar for a complete lists of my published works in various areas.
I regularly participate in different AI/Speech conferences (Neurips, ICML, ICASSP, Interspeech, ICLR, to mention a few) and journals (IEEE TASLP, IEEE SPL, IEEE TSP, Neural Networks, TMLR) in various roles - as Organizer/Reviewer/Program Committee Member/Editor. I am currently serving as an Associate Editor for IEEE Signal Processing Letters and I am also on the Technical Committee of IEEE AASP.
News
- May 2024: Co-organizing URGENT Challenge on Speech Enhancement at Neurips 2024.
- April 2024: Serving as Associate Editor for the journal IEEE Signal Processing Letters.
- Feb 2024: 1 paper accepted at CVPR 2024. Paper on a new multimodal dataset for Room Acoustics and RIR Generation.
- Jan 2024: 1 paper accepted at IEEE TASLP Journal. Paper on a hybrid (DSP + ML) approach for DoA Estimation.
- Dec 2023: 3 papers accepted at ICASSP 2024. Paper on speech enhancement, audiovisual speaker separation, ambisonics spatial audio.
- Nov 2023: Elected to IEEE AASP Technical Committee.
- Oct 2023: PyTorch 2.1 releases Torchaudio-Squim. ICASSP 2023 paper and tutorial code on Torchaudio-Squim.
- Sep 2023: AV-NeRF paper accepted at Neurips 2023.
- Sep 2023: Torchaudio 2.1 paper accepted at IEEE ASRU 2023.
- Jun 2023: 3 papers accepted at Interspeech 2023.
- March 2023: Gave Invited Talk at Carnegie Mellon University on Unsupervised Speech Enhancement and Speech Assessment.
- March 2023: Gave Invited Talk at University of Rochester on Unsupervised Speech Enhancement and Speech Assessment.
- Feb 2023: 1 paper accepted at CVPR 2023. Paper on Egocentric Audio-Visual Object Localization.
- Feb 2023: 6 papers accepted at ICASSP 2023. Papers on Speech Assessment, Single Channel Speech Enhancement, and Audio-Visual Speech Enhancement.
- Sep 2022: Gave Tutorial at Interspeech 2022 on Learning from Weak Labels. Link to Slides
- Jun 2022: Journal Paper accepted to IEEE JSTSP! New work on self-supervised approach for enhancement which can use real-world noisy speech.
- Jun 2022: Ego4D paper selected as a finalist for best paper award at CVPR 2022 !!!
Selected Publications
Google Scholar lists all of my publications.
‡ indicates equal contribution.
Real Acoustic Fields: An Audio-Visual Room Acoustics Dataset and Benchmark
Ziyang Chen, Israel D Gebru, Christian Richardt, Anurag Kumar, William Laney, Andrew Owens, Alexander Richard
IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024.
AV-NeRF: Learning Neural Fields for Real-World Audio-Visual Scene Synthesis
Susan Liang, Chao Huang, Yapeng Tian, Anurag Kumar, Chenliang Xu
Advances in neural information processing systems (Neurips), 2023.
TorchAudio 2.1: Advancing speech recognition, self-supervised learning, and audio processing components for PyTorch
Jeff Hwang, Moto Hira, Caroline Chen, Xiaohui Zhang, Zhaoheng Ni, Guangzhi Sun, Pingchuan Ma, Ruizhe Huang, Vineel Pratap, Yuekai Zhang, Anurag Kumar, Chin-Yun Yu, Chuang Zhu, Chunxi Liu, Jacob Kahn, Mirco Ravanelli, Peng Sun, Shinji Watanabe, Yangyang Shi, Yumeng Tao
2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU).
Torchaudio-Squim: Reference-Less Speech Quality and Intelligibility Measures in Torchaudio
Anurag Kumar, Ke Tan, Zhaoheng Ni, Pranay Manocha, Xiaohui Zhang, Ethan Henderson, Buye Xu
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023.
Egocentric Audio-Visual Object Localization
Chao Huang, Yapeng Tian, Anurag Kumar, Chenliang Xu
IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023.
Remixit: Continual self-training of speech enhancement models via bootstrapped remixing
Efthymios Tzinis, Yossi Adi, Vamsi K Ithapu, Buye Xu, Paris Smaragdis, Anurag Kumar
IEEE Journal of Selected Topics in Signal Processing, 2022.
NORESQA--A Framework for Speech Quality Assessment using Non-Matching References
Pranay Manocha, Buye Xu, Anurag Kumar
Advances in neural information processing systems (Neurips), 2021.
A Sequential Self Teaching Approach for Improving Generalization in Sound Event Recognition
Anurag Kumar, Vamsi Krishna Ithapu
International Conference on Machine Learning (ICML), 2020.
Large Scale Audiovisual Learning of Sounds with Weakly Labeled Data
Haytham Fayek ‡, Anurag Kumar ‡
International Joint Conference on Artificial Intelligence (IJCAI), 2020.
Knowledge Transfer from Weakly Labeled Audio using Convolutional Neural Network for Sound Events and Scenes
Anurag Kumar, Maksim Khadkevich, Christian Fügen
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018.
Audio Event Detection using Weakly Labeled Data
Anurag Kumar, Bhiksha Raj
ACM International Conference on Multimedia (ACM MM), 2016.
Vitæ
Here is my CV (probably approximately correct!).