Chunhui Liu

Staff ML Engineer · TikTok Video Search

I'm a Staff Machine Learning Engineer at TikTok's Video Search Team, where I serve as tech lead for relevance and pretraining. I work with a group of talented individuals to build the world's largest short video search engine with tens of billions of videos, serving billions of users daily. My team focuses on developing BERT/LLM models as part of the search engine, utilizing advanced techniques in CV/NLP/Multimodal learning and pretraining.

Previously, I was an Applied Scientist at Amazon AI, conducting cutting-edge research and developing real-world applications for video and action understanding. I hold a Master's degree in Computer Vision from CMU and a Bachelor's degree in Computer Science, Summa Cum Laude, from Peking University, under the supervision of Prof. Jiaying Liu.

Research & Publications

I remain deeply interested in deep learning, computer vision, multi-modality learning, and video understanding — passionate about building models that understand the physical world the way humans do.

CVPR 2022

TubeR: Tubelet Transformer for Video Action Detection

Jiaojiao Zhao, Yanyi Zhang, Xinyu Li, Hao Chen, Bing Shuai, Mingze Xu, Chunhui Liu, Kaustav Kundu, Yuanjun Xiong, Davide Modolo, Ivan Marsic, Cees GM Snoek, Joseph Tighe

Code Paper

The first SOTA transformer model for action detection, using learnable queries as tubelet proposals.

ICCV 2021

Selective Feature Compression for Efficient Activity Recognition Inference

Chunhui Liu, Xinyu Li, Hao Chen, Davide Modolo, Joseph Tighe

Paper

Using transformers as spatial feature samplers to achieve 6× faster inference without performance drop.

ICCV 2021

VidTr: Video Transformer Without Convolutions

Yanyi Zhang, Xinyu Li, Chunhui Liu, Bing Shuai, Yi Zhu, Biagio Brattoli, Hao Chen, Ivan Marsic, Joseph Tighe

Code Paper

One of the earliest works to apply the transformer architecture for action recognition.

arXiv 2021

A Comprehensive Study of Deep Video Action Recognition

Yi Zhu, Xinyu Li, Chunhui Liu, Mohammadreza Zolfaghari, Yuanjun Xiong, Chongruo Wu, Zhi Zhang, Joseph Tighe, R Manmatha, Mu Li

Code Paper Tutorial

A survey covering 16 datasets and 200+ papers on action understanding, with a full codebase and tutorial workshops.

ACM MM Workshop 2017

PKU-MMD: A Large Scale Benchmark for Continuous Multi-Modal Human Action Understanding

Chunhui Liu, Yueyu Hu, Yanghao Li, Sijie Song, Jiaying Liu

Code Paper Project Page Dataset

A skeleton-based action detection dataset with continuous multi-modal recordings.

Chunhui Liu

Research & Publications

Blog