Kaihua Tang

About Me

I’m a full-time senoir algorithm engineer and a spare-time open-sourcerer. My current work mainly includes on-device LLM deployment and inference speedup. I also have experience in computer vision research and game development. My lifelong dream is to empower video games with more intelligence in the gameplay.

I received my Ph.D. degree from the School of Computer Science and Engineering at Nanyang Technological University in 2021, dual-master degrees in Computer Science from the joint programme of Shanghai Jiao Tong University and Waseda University in 2018, and B.E. degree in Computer Science from the IEEE Pilot Class at Shanghai Jiao Tong University (SJTU) in 2015.

News

[Mar. 2024] I received the Outstanding New Employee Award 2023 & Future Star Award 2023 of SRC.
[Oct. 2023] Our developed on-device LLM was released at the China Mobile Global Partners Conference 2023.
[Sep. 2023] One Paper was accepted to NeurIPS 2023.
[Jul. 2022] We are organizing the 2st Causality in Vision Workshop at ECCV 2022.
[Jul. 2022] I was recognized as an outstanding reviewer (Top 10%) for ICML 2022.
[Jul. 2022] Three Papers (One Oral) were accepted to ECCV 2022.
[Dec. 2021] I passed my Ph.D. thesis defense. Thanks for all the helps in the past 3 years.
[May. 2021] I received the Silver Award in 2021 PREMIA Best Student Paper Award.
[Mar. 2021] I received the 2021 Alibaba Outstanding Interns in Academic Cooperation award.
[Mar. 2021] We organized the 1st Causality in Vision Workshop at CVPR 2021.
[Jan. 2021] Two Papers were accepted to CVPR 2021.

Selected Publications [Google Scholar]

ECCV

Invariant Feature Learning for Generalized Long-Tailed Classification

Kaihua Tang, Mingyuan Tao, Jiaxin Qi, Zhenguang Liu, Hanwang Zhang

European Conference on Computer Vision (ECCV), 2022.

PDF Code Project Page BibTex

ECCV

Identifying Hard Noise in Long-Tailed Sample Distribution

Xuanyu Yi, Kaihua Tang, Xian-Sheng Hua, Joo-Hwee Lim, Hanwang Zhang

European Conference on Computer Vision (ECCV), 2022.

PDF Code BibTex Oral Presentation

ECCV

Class Is Invariant to Context and Vice Versa: On Learning Invariance for Out-Of-Distribution Generalization

Jiaxin Qi, Kaihua Tang, Qianru Sun, Xian-Sheng Hua, Hanwang Zhang

European Conference on Computer Vision (ECCV), 2022.

PDF Code BibTex

CVPR

Distilling Causal Effect of Data in Class-Incremental Learning

Xinting Hu, Kaihua Tang, Chunyan Miao, Xian-Sheng Hua, Hanwang Zhang

IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021.

PDF Code BibTex

CVPR

Counterfactual VQA: A Cause-Effect Look at Language Bias

Yulei Niu, Kaihua Tang, Hanwang Zhang, Zhiwu Lu, Xian-Sheng Hua, Ji-Rong Wen

IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021.

PDF Code BibTex

NeurIPS

Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effect

Kaihua Tang, Jianqiang Huang, Hanwang Zhang

Conference on Neural Information Processing Systems (NeurIPS), 2020.

PDF Code BibTex 516 Github Stars • 229 Citations

CVPR

Unbiased Scene Graph Generation from Biased Training

Kaihua Tang, Yulei Niu, Jianqiang Huang, Jiaxin Shi, Hanwang Zhang

IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020.

PDF Code BibTex Oral Presentation • 836 Github Stars • 257 Citations

CVPR

Learning to Segment the Tail

Xinting Hu, Yi Jiang, Kaihua Tang, Hanwang Zhang, Chunyan Miao, Jingyuan Chen

IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020.

PDF Code BibTex

CVPR

Learning to Compose Dynamic Tree Structures for Visual Contexts

Kaihua Tang, Hanwang Zhang, Baoyuan Wu, Wenhan Luo, Wei Liu

IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019.

PDF Code BibTex Best Paper Finalists [45/5160] • Oral Presentation • 280 Citations

CVPR

Auto-Encoding Scene Graphs for Image Captioning

Xu Yang, Kaihua Tang, Hanwang Zhang, Jianfei Cai

IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019.

PDF Code BibTex Oral Presentation • 556 Citations

Awards

Outstanding Reviewer (Top 10%), ICML, 2022
2021 Alibaba Outstanding Interns in Academic Cooperation, Alibaba Group, 2021
2021 & 2019 PREMIA Best Student Paper Award, 2nd Place, PREMIA, 2021 & 2019
CVPR 2019 Best Paper Finalists, 2019
Honorable Judge Award, The 5th Cloud Programming World Cup, FORUM8 Tokyo, 2017
Waseda Partial Tuition-Waiver Scholarship for International Students (10/300), Waseda University, 2015
IPS special scholarship for international students, Waseda University, 2014
Monbukagakusho Honors Scholarship for International Students, JASSO, 2014
Emerging Talent Award, The 1st Cloud Programming World Cup, FORUM8 Tokyo, 2013

Academic Services

Organizing Committees

1st Causality in Vision Workshop at CVPR 2021

2st Causality in Vision Workshop at ECCV 2022

Talks and Blogs

Invited Talk : To TechBeat, Hosted by TechBeat AI Community, Online Sharing, 2022.10

Invited Talk : To Kuaishou, Recommendation System Group, Beijing, 2021.07

Invited Talk : To Fudan University, Department of CS, Hosted by Asso Prof.Jingjing Chen, Shanghai, 2021.07

Invited Talk : To The Lab for Media Search (LMS), NUS, 2021.06

Invited Talk : To VALSE Webinar (volume 20-29), Hosted by Prof.Meng Yang, Online, 2020.12

Invited Talk : To Causal AI Reading Group, Jointly Hosted by Swarma Club and BAAI Hub, 2020.12

Invited Talk : To Jishi Community, Hosted by Jishi Team, BiliBili, 2020.12

Invited Talk : To Alibaba Group, Hosted by Tianchi Team from Alibaba Cloud, Hangzhou, China, 2020.11

Invited Talk : To Jiantao Jiao's Lab, Hosted by Ph.D. Banghua Zhou, UC Berkeley, 2020.10

Blogs : Sharing Research Experiences at Zhihu, Language: Chinese

Paper Review

CVPR, ECCV, ICCV, WACV, NeurIPS, ICLR, ICML, AAAI, TPAMI

Projects

Scene Graph Benchmark in Pytorch (835 Github Stars)

This project aims to provide a new codebase for Scene Graph Generation (SGG). It is built on top of the well-known maskrcnn-benchmark. Moreover, I included all the exsiting metrics: R@K, mR@K, ngR@K, zR@K, to benchmark the SGG.

Long-Tailed Recognition in Pytorch (516 Github Stars)

This project provides a strong single-stage baseline for Long-Tailed Classification, Detection, and Instance Segmentation (under LVIS dataset). This project can be easily generalized to other tasks with unbalanced datasets.

VQA2.0 Recent Approachs 2018 in Pytorch (273 Github Stars)

An open-source visual question answering (VQA) codebase built on top of the bottom-up-attention-vqa. It integrates several popular VQA papers published in 2018.

Indie Game Development

Out of interest, I independently developed several mobile games on Iphone. They have been downloaded over 10k times on Apple store in half a year.

Experience

Alibaba, DAMO Academy, Research Intern (2019.7 - 2021.11)

Tencent, AI Lab, Research Intern (2018.3 - 2018.6)

Mihoyo, Software Engineer Intern (2017.4 - 2017.12)

Mobile Game Development Using Unity 3D.

Toshiba, Research & Development Intern (2015.8 - 2015.9)

Speech Lab Intern, SJTU (2014.3 - 2014.9)

28th ACM-MM Volunteer, Seattle, USA (2020.10)

Received Volunteer Appreciation Certification in the 2020 ACM Multimedia for joining the organization of online presentation.

YAPM Summer Volunteer, Yunnan Province, China (2014.7 - 2014.8)

Youth Ambassador Program for Minorities (TECC Organization) is determined to help the youth generation of minorities in remote area of China to inherit and protect their cultures.