Bo Yang

I am an Assistant Professor (2020.11-) in the Department of Computing at The Hong Kong Polytechnic University. I completed my D.Phil degree (2016.10-2020.09) in the Department of Computer Science at University of Oxford, supervised by Profs. Niki Trigoni and Andrew Markham. Prior to Oxford, I obtained an M.Phil degree from The University of Hong Kong and a B.Eng degree from Beijing University of Posts and Telecommunications.

In my D.Phil study, I interned at the Augumented Reality team of Amazon (Palo Alto, CA). In my M.Phil study, I interned at Hong Kong Applied Science and Technology Research Institute. In my undergraduate study, I was an exchange student at Universitat Politècnica de València (Valencia, Spain).

Email / vLAR Group / Google Scholar / Github/ Xiaohongshu(小红书)

Research

I lead the Visual Learning and Reasoning (vLAR) Group, focusing on the fundamental research problems in machine learning, computer vision, and robotics. Our research goal is to build intelligent systems which endow machines to recover, understand, and eventually interact with the real 3D world. This includes accurate and efficient recognition, segmentation and reconstruction of all individual objects within large-scale 3D scenes.


Openings (Jul 2025 - ):

Several fully funded PostDoc/PhD/RA positions for 3D vision and robot learning are available now. Email me with your CV and transcripts!

(All emails/CVs are carefully read and evaluated. Only matched candidates will be responded.)

News & Activities

[2025.06.25] Our TRACE/RayletDF (Highlight) on 3D physics/geometry learning are accepted by ICCV 2025.

[2025.05.01] Our paper unMORE on object-centric learning is accepted by ICML 2025.

[2025.02.27] Our papers FreeGave/LogoSP on 3D physics/semantics learning are accepted by CVPR 2025.

[2025.01.23] Our paper GrabS (Spotlight) on 3D semantics learning is accepted by ICLR 2025.

[2024.05.23] Our extension of OGC on 3D semantics learning is accepted by TPAMI.

[2024.05.02] Our paper OSN on 3D geometry learning is accepted by ICML 2024.

[2024.01.29] Our paper DynCatch on robot learning is accepted by ICRA 2024.

[2024.01.09] Our extension of UnsupObjSeg on object-centric learning is accepted by IJCV.

[2023.09.22] Our papers RayDF/NVFi on 3D geometry/physics learning are accepted by NeurIPS 2023.

[2023.03.31] We are going to organize The 3rd Challenge on Point Cloud Understanding at ICCV 2023.

[2023.02.28] Our paper GrowSP on 3D semantics learning is accepted by CVPR 2023.

[2023.01.22] Our paper DM-NeRF on 3D semantics learning is accepted by ICLR 2023.

[2023.01.18] Our paper DecoupSkill on robot learning is accepted by ICRA 2023.

[2022.09.03] Our papers OGC/UnsupObjSeg on semantics learning are accepted by NeurIPS 2022.

[2022.07.03] Our paper SQN on 3D semantics learning is accepted by ECCV 2022.

[2022.05.26] Our extension of SpinNet on 3D geometry learning is accepted by TPAMI.

[2022.05.04] We are going to organize The 2nd Challenge on Point Cloud Understanding at ECCV 2022.

[2021.11.11] Our extension of SensatUrban on 3D semantic learning is accepted by IJCV.

[2021.07.23] Our paper GRF on 3D geometry learning is accepted by ICCV 2021.

[2021.05.15] Our extension of RandLA-Net on 3D semantic learning is accepted by TPAMI.

[2021.04.09] We are going to organize The 1st Challenge on Point Cloud Understanding at ICCV 2021.

[2021.03.01] Our SensatUrban/SpinNet on 3D semantics/geometry learning are accepted by CVPR 2021.

[2021.02.28] Our paper RadarLoc is accepted by ICRA 2021.

[2020.11.28] We organize a tutorial about 3D point cloud learning at 3DV 2020.

[2020.02.27] Our paper RandLA-Net (Oral) on 3D semantics learning is accepted by CVPR 2020.

[2019.09.03] Our paper 3D-BoNet (Spotlight) on 3D semantics learning is accepted by NeurIPS 2019.

[2019.08.16] Our paper AttSets on 3D geometry learning is accepted by IJCV.

[2018.08.22] Our paper 3D-RecGAN++ on 3D geometry learning is accepted by TPAMI.

Five Selected Recent Publications (Full list at vLAR research page)
PontTuset

OSN: Infinite Representations of Dynamic 3D Scenes from Monocular Videos
Z. Song, J. Li, B. Yang
International Conference on Machine Learning (ICML), 2024
arXiv / Code

We present the first framework to represent dynamic 3D scenes in infinitely many ways from a monocular RGB video.

PontTuset

RayDF: Neural Ray-surface Distance Fields with Multi-view Consistency
Z. Liu, B. Yang*, Y. Luximon, A. Kumar, J. Li
Advances in Neural Information Processing Systems (NeurIPS), 2023
arXiv / Project Page / Code
(* indicates corresponding author)

We propose a novel ray-based 3D shape representation, achieving a 1000x faster speed in rendering.

PontTuset

NVFi: Neural Velocity Fields for 3D Physics Learning from Dynamic Videos
J. Li, Z. Song, B. Yang
Advances in Neural Information Processing Systems (NeurIPS), 2023
arXiv / Code

We present a novel framework to simultaneously learn the geometry, appearance, and physical velocity of 3D scenes.

PontTuset

GrowSP: Unsupervised Semantic Segmentation of 3D Point Clouds
Z. Zhang, B. Yang*, B. Wang, B. Li
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023
arXiv / Code
(* indicates corresponding author)

We propose the first unsupervised 3D semantic segmentation method, learning from growing superpoints in point clouds.

PontTuset

OGC: Unsupervised 3D Object Segmentation from Rigid Dynamics of Point Clouds
Z. Song, B. Yang
Advances in Neural Information Processing Systems (NeurIPS), 2022
arXiv / Video/ Code

We introduce the first unsupervised 3D object segmentation method on point clouds.


DPhil (PhD) Thesis
PontTuset

Learning to Reconstruct and Segment 3D Objects

B. Yang
Oxford Research Archive (PDF), 2020; (News: 机器之心报道)

Thesis committee (Transfer/Confirmation/Viva):
Alessandro Abate, Andrew Davison, Pawan Kumar, Andrew Zisserman.

This thesis aims to understand scenes and the objects within them by learning general and robust representations using deep neural networks, trained on large-scale real-world 3D data. In particular, the thesis makes three core contributions from object-level 3D shape estimation from single or multiple views to scene-level semantic understanding.

Talks & Services

[2025.07] Invited talk about 3D Physics Learning at Chaspark Live (Video and Transcript).

[2025.06] Invited talk about Unsupervised 3D Spatial Understanding of Point Clouds at MMT 2025.

[2025.04] Invited talk about Unsupervised 3D Semantics Learning at Cambridge University.

[2025.04] Invited talk about 3D Physics Learning at CVM 2025.

[2024.12] Invited talk about 3D Physics and Semantics Learning at Tongji University.

[2023.05] Invited talk about Unsupervised 3D Semantic and Instance Segmentation at VALSE webinar (Video).

[2022.12] Invited talk about Unsupervised 2D/3D Object Segmentation at TechBeat forum (Video).

[2022.06] Invited talk about 3D Scene Reconstruction, Decomposition and Manipulation at Xiamen University.

[2021.10] Invited talk about 3D Representation Learning at GAMES Webinar (Video).

[2021.04] Invited talk about Beyond Supervised Learning for 3D Representations at a CSIG workshop (Video).

[2020.10] Invited talk about 3D Scene Understanding at Wonderland AI Summit (Video).

[2020.09] Invited talk about 3D Point Cloud Segmentation at MFI 2020.

[2020.03] Invited talk about our RandLA-Net and 3D-BoNet at Shenlan (Video and Slides).

[2018 -] Regularly chairing/reviewing for top-tier conferences/journals in ML, CV, and robotics.

Teaching

Spring, 2024&2025: AI and Big Data Computing in Practice (The Hong Kong Polytechnic University).

Fall, 2023&2024&2025: Machine Learning and Data Analytics (The Hong Kong Polytechnic University).

Spring, 2023&2024&2025: Creative Digital Media Design (The Hong Kong Polytechnic University).

Spring&Fall, 2021&2022: Machine Learning and Data Analytics (The Hong Kong Polytechnic University).

Hilary, 2019: Knowledge Representation & Reasoning (University of Oxford).

Michaelmas, 2018&2017: Machine Learning (University of Oxford).

Spring, 2014: C++ Programming (The University of Hong Kong).

Mentoring (Full list at vLAR member page)

Qingyong Hu (Oct 2018 - ): Department of Computer Science at University of Oxford.

Alexander Trevithick (Oct 2019 - Mar 2021): Now PhD at UCSD.

Jianan Wang (May - Dec 2018): Now with Google DeepMind.

Zihang Lai (Oct 2017 - Mar 2018): Now PhD at CMU.

About Me

In my free time, I like playing tennis on lawns, clays, and hard surfaces. I also like to fly drones for landscape photography. Here's a video over the historic Oxford [Youtube, 腾讯视频], and another video for the scenic Lake District [Youtube]. Remember to turn up the volume for the background music.


Last update: 2025.07. Thanks.