Guosheng Lin

Guosheng Lin is a computer vision researcher at Nanyang Technological University in Singapore whose work covers semantic segmentation, few-shot learning and 3D deep learning. He studied at South China University of Technology, earned a PhD at the University of Adelaide in 2014 with a Google Australia Fellowship, and then worked as a research fellow at the Australian Centre for Robotic Vision. In 2017 he presented RefineNet, a multi-path refinement network for high-resolution semantic segmentation, and joined NTU as an Assistant Professor. He became an Associate Professor with tenure in 2023, and since 2025 he is Assistant Dean (Graduate Studies) of NTU's College of Computing and Data Science.

Now at: Nanyang Technological University

Timeline

2003–2014

Degrees in Guangzhou and Adelaide

He completes a B.Eng in 2007 and an M.Eng in 2010 at South China University of Technology in Guangzhou. He then studies for a PhD at the University of Adelaide, supervised by Chunhua Shen with David Suter and Tat-Jun Chin, and completes it in 2014.
2014

A Google Australia fellowship

In 2014, he receives the Google Australia Fellowship in Machine Perception Research for his work on binary code learning with deep neural networks for image retrieval. The University of Adelaide also gives his thesis a Dean's Commendation for Doctoral Thesis Excellence in 2014.
2014–2017

Research fellow at ACRV

From 2014 to 2017, he is a research fellow at the Australian Centre for Robotic Vision at the University of Adelaide, working closely with Chunhua Shen, Ian Reid and Anton van den Hengel. In 2016, he coauthors a CVPR paper on efficient piecewise training of deep structured models for semantic segmentation, which reaches an intersection-over-union score of 78.0 on PASCAL VOC 2012.
2017

RefineNet for segmentation

At CVPR 2017, he presents RefineNet with Anton Milan, Chunhua Shen and Ian Reid. The multi-path refinement network reaches an intersection-over-union score of 83.4 on PASCAL VOC 2012.
2017–2023

Assistant Professor at NTU

He joins Nanyang Technological University in Singapore as an Assistant Professor in 2017. His research there covers computer vision, scene understanding, 3D deep learning and generative AI.
2019–2020

Few-shot segmentation and classification

In 2019, he coauthors CANet, a class-agnostic network that segments new classes from a few annotated images and reaches 55.4% mean intersection-over-union for 1-shot segmentation on PASCAL VOC 2012. In 2020, he coauthors DeepEMD, which uses the Earth Mover's Distance for few-shot image classification.
2023

Tenured Associate Professor

NTU promotes him to Associate Professor with tenure in 2023. He is also an Associate Editor of IEEE Transactions on Circuits and Systems for Video Technology from 2023.
Since 2025

Assistant Dean for graduate studies

In 2025, he becomes Assistant Dean (Graduate Studies) of the College of Computing and Data Science at NTU. He is also an Area Chair for CVPR 2024 and 2025.