Bo Yang

I am a D.Phil student (Oct 2016 - ) 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 where I was supervised by Prof. S.H. Choi, and a B.Eng degree from Beijing University of Posts and Telecommunications.

In this summer (July - Oct 2019), I was a research intern 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 / Google Scholar  /  Github

I'm interested in machine learning, computer vision, and robotics. My 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.

[2019.11.27] One co-authored paper on 3D semantic segmentation is on arXiv .

[2019.10.24] Successfully defend D.Phil confirmation viva, examined by Profs. Andrew Zisserman and Alessandro Abate.

[2019.09.03] One first-authored paper on 3D instance segmentation is accepted as a spotlight at NeurIPS 2019.

[2019.08.16] One first-authored paper on multi-view 3D reconstruction is accepted in IJCV.

RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds
Q. Hu, B. Yang*, L. Xie, S. Rosa, Y. Guo, Z. Wang, N. Trigoni, A. Markham
arXiv / Code / Video / Semantic3D Benchmark / SemanticKITTI Benchmark

We introduce an efficient and lightweight neural architecture to directly infer per-point semantics for large-scale point clouds.

Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds
B. Yang, J. Wang, R. Clark, Q. Hu, S. Wang, A. Markham, N. Trigoni
Conference on Neural Information Processing Systems (NeurIPS), 2019 (Spotlight, 200/6743)
arXiv / Code / ScanNet Benchmark / Reddit Discussion

We propose a simple and efficient neural architecture for accurate 3D instance segmentation on point clouds. It achieves the SOTA performance on ScanNet and S3DIS (June 2019).

DeepPCO: End-to-End Point Cloud Odometry through Deep Parallel Neural Network
W. Wang, M.R.U. Saputra, P. Zhao, P. Gusmao, B. Yang, C. Chen, A. Markham, N. Trigoni
EEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2019
arXiv

We propose a novel end-to-end deep parallel neural network to estimate the 6-DOF poses using consecutive 3D point clouds.

Robust Attentional Aggregation of Deep Feature Sets for Multi-view 3D Reconstruction
B. Yang, S. Wang, A. Markham, N. Trigoni
International Journal of Computer Vision (IJCV), 2019 (IF=6.07)
arXiv / Code / Springer Open Access

We propose an attentive aggregation module together with a training algorithm for multi-view 3D object reconstruction. It outperforms all existing poolings and recurrent neural networks.

Learning Semantically Meaningful Embeddings Using Linear Constraints
S. Lin, B. Yang, R. Birke, R. Clark
IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR-W), 2019
CVF Open Access

We propose a simple embedding learning method that jointly optimises for an auto-encoding reconstruction task and for estimating the corresponding attribute labels.

Dense 3D Object Reconstruction from a Single Depth View
B. Yang, S. Rosa, A. Markham, N. Trigoni, H. Wen
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2018 (IF=17.73)
arXiv / Code / IEEE Xplore

We propose a novel neural architecture to reconstruct the complete 3D structure of a given object from a single arbitrary depth view using generative adversarial networks.

3D-PhysNet: Learning the Intuitive Physics of Non-Rigid Object Deformations
Z. Wang, S. Rosa, B. Yang, S. Wang, N. Trigoni, A. Markham
International Joint Conference on Artificial Intelligence (IJCAI), 2018
arXiv / Code

We present a neural framework to predict how a 3D object will deform under an applied force using intuitive physics modelling.

Learning 3D Scene Semantics and Structure from a Single Depth Image
B. Yang, Z. Lai, X. Lu, S. Lin, H. Wen, A. Markham, N. Trigoni
IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR-W), 2018
CVF Open Access / IEEE Xplore

We propose an efficient and holistic pipeline to simultaneously learn the semantics and structure of a scene from a single depth image.

Defo-Net: Learning Body Deformation Using Generative Adversarial Networks
Z. Wang, S. Rosa, L. Xie, B. Yang, S. Wang, N. Trigoni, A. Markham
IEEE International Conference on Robotics and Automation (ICRA) , 2018
arXiv / Code / Video/ IEEE Xplore

We present a novel generative adversarial network to predict body deformations under external forces from a single RGB-D image.

3D Object Reconstruction from a Single Depth View with Adversarial Learning
B. Yang, H. Wen, S. Wang, R. Clark, A. Markham, N. Trigoni
IEEE International Conference on Computer Vision Workshops (ICCV-W) , 2017
arXiv / Code / IEEE Xplore/ News(机器之心报道)

We propose a novel approach to reconstruct the complete 3D structure of a given object from a single arbitrary depth view using generative adversarial networks.

Updating Wireless Signal Map with Bayesian Compressive Sensing
B. Yang, S. He, S-H G. Chan
ACM International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWiM) , 2016
ACM DL

We propose Compressive Signal Reconstruction (CSR), a novel learning system employing Bayesian compressive sensing (BCS) for online signal map update.

A mechanised 3D scanning method for item-level radio frequency identification of palletised products
S.H. Choi, B. Yang, H.H. Cheung
Computers in Industry , 2015 (IF=4.77)
Elsevier ScienceDirect

We propose a mechanised 3D scanning method for identification of tagged products in large numbers to facilitate supply chain management.

Item-level RFID for Enhancement of Customer Shopping Experience in Apparel Retail
S.H. Choi, Y.X. Yang, B. Yang, H.H. Cheung
Computers in Industry , 2015 (IF=4.77)
Elsevier ScienceDirect

We propose an item-level RFID-enabled retail store management system for relatively high-end apparel products to provide customers with more leisure, interaction for product information.

RFID Tag Data Processing in Manufacturing for Track-and-Trace Anti-counterfeiting
S.H. Choi, B. Yang, H.H. Cheung, Y.X. Yang
Computers in Industry , 2015 (IF=4.77)
Elsevier ScienceDirect

We present a track-and-trace anti-counterfeiting system, and propose a tag data processing and synchronization algorithm to generate initial e-pedigrees for products.

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

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

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

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

Alexander Trevithick (Oct 2019 - ):    Exeter College at University of Oxford.

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

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

Zihang Lai (Oct 2017 - Mar 2018):    Now with VGG.

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: 2019.11.27. Thanks.