Hengshuang Zhao

Hengshuang Zhao is a computer vision researcher whose models help machines parse scenes, 3D point clouds and depth from single images. Early in his career, he built the Pyramid Scene Parsing Network (PSPNet) with Jiaya Jia and colleagues; it came first in the ImageNet scene parsing challenge in 2016 and has gathered over 20,000 citations. After a CUHK doctorate and postdoctoral posts at Oxford's Torr Vision Group and MIT CSAIL, he coauthored Point Transformer in 2021 and Depth Anything in 2024, a depth model integrated into Apple Core ML Models in June 2024. China's National Natural Science Foundation gave him an Excellent Young Scientists Fund award in August 2024, and in August 2026 HKU announced his place on MIT Technology Review's 2025 '35 Innovators Under 35 China' list. He is an assistant professor at the University of Hong Kong's School of Computing and Data Science, where he directs the Super Artificial Intelligence Lab (SAIL).

Now at: University of Hong Kong
Badges: Bridge

Timeline

2016–2017

The PSPNet scene parser

He coauthors the Pyramid Scene Parsing Network (PSPNet) with Jianping Shi, Xiaojuan Qi, Xiaogang Wang and Jiaya Jia; its pyramid pooling module gathers global context, and it comes first in the ImageNet scene parsing challenge in 2016. The paper appears at CVPR 2017, and PSPNet goes on to collect over 20,000 citations as a framework for semantic segmentation.
2017–2018

A CUHK doctorate and internships

After a bachelor's degree at Huazhong University of Science and Technology, he earns a PhD at the Chinese University of Hong Kong, supervised by Jiaya Jia. During his doctorate, he interns at Adobe Research's Creative Intelligence Lab in San Jose in 2017 and with Raquel Urtasun at Uber's Advanced Technologies Group in Toronto in 2018.
2019–2022

Postdocs at Oxford and MIT

From 2019 to 2020, he works with Vladlen Koltun at Intel's Intelligent Systems Lab in Santa Clara as a research intern and visiting researcher. He is then a postdoctoral researcher in Philip Torr's Torr Vision Group at Oxford from 2020 to 2021 and at MIT's Computer Science and Artificial Intelligence Laboratory under Antonio Torralba from 2021 to 2022.
2021

Point Transformer for 3D data

In 2021, he presents Point Transformer at ICCV with Li Jiang, Jiaya Jia, Philip Torr and Vladlen Koltun. The network applies self-attention layers to 3D point clouds and reaches 70.4% mIoU on Area 5 of the S3DIS segmentation dataset.
2024

Depth Anything

He coauthors Depth Anything in 2024 with Lihe Yang and others, a foundation model for monocular depth estimation whose data engine automatically annotates about 62 million unlabeled samples. In June 2024, Depth Anything wins a CVPR 2024 Best Demo Honorable Mention and is integrated into Apple Core ML Models.
August 2024

An Excellent Young Scientists award

In August 2024, he receives the Excellent Young Scientists Fund from the National Natural Science Foundation of China.
August 2026

MIT Technology Review TR35 China

In August 2026, HKU announces his selection for MIT Technology Review's 2025 '35 Innovators Under 35 China' list in the Pioneers category. The honour recognizes his physical intelligence work, including PSPNet, the Point Transformer series, Depth Anything, and systems such as DriveGPT4 and GPT4Point.
As of October 2026

Professor at HKU

He is an assistant professor at the University of Hong Kong's School of Computing and Data Science, where he directs the Super Artificial Intelligence Lab (SAIL). His group works on embodied, multimodal, spatial and generative AI.