pottrait

Hanxue (Charles) Liang

University of Cambridge

EmailGoogle ScholarLinkedin

About Me

I am a final-year PhD student in Computer Science at University of Cambridge, advised by Prof. Cengiz Öztireli and Prof. Rafal Mantiuk. Before starting my Ph.D., I obtained my Master's degree in Robotics, Systems and Control from ETH Zurich under supervision of Prof. Luc Van Gool.
I collaborate closely with Dr. Jiahui Huang, Dr. Zan Gojcic and Prof. Sanja Fidler on 3D/4D Reconstruction and Generation; with Prof. Atlas Wang and Prof. Yu Cheng on topic of on-device Mixture of Expert for Multi-Task/Modality Learning and Generalizable Novel View Synthesis; with Dr. Fangneng Zhan and Prof. Christian Theobalt on Evolutive Rendering Optimization; with Prof. Luc Van Gool on Self-supervised Learning for Point Cloud Understanding. I worked as research scientist intern at Nvidia, Qualcomm, Google Research and Microsoft Research. I was the finalist of Qualcomm Innovation Fellowship 2024.

My research focuses on computer vision, graphics and machine learning, with particular expertise in:

Recent News

  • Our work αSurf is accepted by 3DV'25 as oral presentation.
  • Our work Diffusion4D, L4GM, SCube are accepted by NeurIPS'24. Congratulations to our collaborators.
  • I was one of the finalists of the Qualcomm Innovation Fellowship 2024 (QIF 2024). Innovation title: "Real-time On-device 3D Modeling via Neural Implicit Scene Representations".
  • Our work NeRF-NQA is accepted by Transactions on Visualization and Computer Graphics (TVCG) 24.
  • Our work Perceptual-NeRF is accepted by EUROGRAPHICS'24.
  • Our work GNT-MOVE is accepted by ICCV'23.
  • We won 3rd place in the University Demo Best Demonstration at the 59th Design Automation Conference (DAC 2022). We demo for a multi-task vision transformer on FPGA.
  • Our work M3ViT is accepted by NeurIPS'22.
  • Our work MLSP is accepted by ECCV'22.

Research Overview

Research Overview Image

Selected Publications
Full publication list at Google Scholar

* denotes equal contribution, † denotes project lead.

3D/4D Reconstruction and Generation from Images/Monocular Video

Feed-Forward Bullet-Time Reconstruction of Dynamic Scenes from Monocular Videos
Hanxue Liang*†Jiawei Ren*Ashkan Mirzaei*Antonio TorralbaZiwei LiuIgor GilitschenskiSanja FidlerCengiz OztireliHuan LingZan GojcicJiahui Huang
Arxiv, 2024 
PaperProject
Evolutive Rendering Models
Hanxue Liang*Fangneng Zhan*Yifan WangMichael NiemeyerMichael OechsleAdam KortylewskiCengiz OztireliGordon WetzsteinChristian Theobalt
Arxiv, 2024 
PaperProject
Perceptual Quality Assessment of NeRF and Neural View Synthesis Methods for Front-Facing Views
Hanxue LiangTianhao WuParam HanjiFrancesco BanterleHongyun GaoRafał MantiukCengiz Oztireli
EUROGRAPHICS 2024 
PaperProject
GNT-MOVE framework loading...
Enhancing NeRF akin to Enhancing LLMs: Generalizable NeRF Transformer with Mixture-of-View-Experts
Hanxue Liang*†Wenyan Cong*Peihao WangZhiwen FanTianlong ChenMukund VarmaYi WangZhangyang Wang
ICCV2023 
PaperCode
αSurf: Implicit Surface Reconstruction for Semi-Transparent and Thin Objects with Decoupled Geometry and Opacity
Tianhao WuHanxue LiangFangcheng ZhongGernot RieglerShimon VainerJiankang DengCengiz Oztireli
3DV 2025  (Oral Presentation)
PaperProjectCode
Diffusion4D: Fast Spatial-temporal Consistent 4D Generation via Video Diffusion Models
Hanwen Liang*Yuyang Yin*Dejia XuHanxue LiangZhangyang WangKonstantinos N. PlataniotisYao ZhaoYunchao Wei
NeurIPS 2024 
PaperProjectCode
L4GM: Large 4D Gaussian Reconstruction Model
Jiawei RenKevin XieAshkan MirzaeiHanxue LiangXiaohui ZengKarsten KreisZiwei LiuAntonio TorralbaSanja FidlerSeung Wook KimHuan Ling
NeurIPS 2024 
PaperProjectCode
SCube: Instant Large-Scale Scene Reconstruction using VoxSplats
Xuanchi Ren*Yifan Lu*Hanxue LiangZhangjie WuHuan LingMike ChenSanja FidlerFrancis WilliamsJiahui Huang
NeurIPS 2024 
PaperProjectCode
GNT-MOVE framework loading...
NeRF-NQA: No-Reference Quality Assessment for Scenes Generated by NeRF and Neural View Synthesis Methods
Qiang QuHanxue LiangXiaoming ChenYuk Ying ChungYiran Shen
TVCG2024 
PaperCode

On-device Mixture of Experts for Multi-task/modality Learning Agent

M^3ViT Video thumbnail loading...
M3ViT: Mixture-of-Experts Vision Transformer for Efficient Multi-task Learning with Model-Accelerator Co-design
Hanxue Liang*†Zhiwen Fan*Rishov SarkarZiyu JiangTianlong ChenKai ZouYu ChengCong HaoZhangyang Wang
NeurIPS2022  (QIF 2022 Award & DAC 3rd best demo)
PaperCode
M^3ViT Video thumbnail loading...
Edge-MoE: Memory-Efficient Multi-Task Vision Transformer Architecture with Task-level Sparsity via Mixture-of-Experts
Rishov SarkarHanxue LiangZhiwen FanZhangyang WangCong Hao
ICCAD2023 
PaperCode

Self-supervised Learning for Point Cloud Understanding

MLSP image loading...
Point Cloud Domain Adaptation via Masked Local 3D Structure Prediction
Hanxue LiangHehe FanZhiwen FanYi WangTianlong ChenYu ChengZhangyang Wang
ECCV2022 
PaperCode
M^3ViT Video thumbnail loading...
Exploring Geometry-aware Contrast and Clustering Harmonization for Self-supervised 3D Object Detection
Hanxue Liang*Chenhan Jiang*Dapeng FengXin ChenHang XuXiaodan LiangWei ZhangZhenguo LiLuc Van Gool
ICCV2021 
Paper
M^3ViT Video thumbnail loading...
One Million Scenes for Autonomous Driving:ONCE Dataset
Jiageng MaoMinzhe NiuChenhan JiangHanxue LiangJingheng ChenXiaodan LiangYamin LiChaoqiang YeWei ZhangZhenguo LiJie YuHang XuChunjing Xu
NeurIPS Datatrack2021 
PaperProject

Services

  • Conference Reviewers: 
    NeurIPS 23/24, ICLR 24/25, CVPR 24/25, ECCV 24 .

Teaching and Supervision (Uni of Cam)

  •  
    Further Graphics 2023-2024
  •  
    Machine Vision Perception 2024-2025
  •  
    Further Graphics 2024-2025