AOP-Net: All-in-One Perception Network for Joint LiDAR-based 3D Object Detection and Panoptic Segmentation
Yixuan Xu*, Hamidreza Fazlali*, Bingbing Liu, Yuan Ren
to appear at 2023 IEEE Intelligent Vehicles Symposium (IV)
Hi, thank you for paying a visit!
I am a student researcher fascinated by innovative technologies and machine intelligence. I finished three years of study as an Engineering Science student at the University of Toronto, majoring in Machine Intelligence. Currently on a gap year, I am fortunate to be taking part in the development of autonomous driving technologies at Huawei Noah's Ark Research Lab. Specifically, I conduct research on 3D perception with LiDAR point clouds. In the past, I took multiple leadership roles in diverse fields. I cherish the moments when I served as a military instructor at the Singapore Armed Forces, as a co-founder for a fitness tech startup, OpenRace, and recently as an advocate building a machine intelligence student community, UTMIST. I find it rewarding to serve my community, tackle challenges, and deliver beneficial goods. I am always looking to build more meaningful things.
Education
Candidate for the Bachelor's Degree of Applied Science in Engineering Science, Machine Intelligence
University of Toronto
September 2018 - April 2023 | Toronto, ON
Dean's List. Minor in Engineering Business. Co-president at University of Toronto Machine Intelligence Student Team (UTMIST).
Research
I am curious about machine learning, computer vision, and generation of images and audio. My research works so far involve LiDAR-based 3D object detection and segmentation.
CPSeg: Cluster-free Panoptic Segmentation of 3D LiDAR Point Clouds
Enxu Li*, Ryan Razani*, Yixuan Xu, Bingbing Liu
to appear at 2023 IEEE International Conference on Robotics and Automation (ICRA)
A Versatile Multi-View Framework for LiDAR-based 3D Object Detection with Guidance from Panoptic Segmentation
Hamidreza Fazlali, Yixuan Xu, Yuan Ren, Bingbing Liu
presented at 2022 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
SMAC-Seg: LiDAR Panoptic Segmentation via Sparse Multi-directional Attention Clustering
Enxu Li*, Ryan Razani*, Yixuan Xu, Bingbing Liu
presented at 2022 IEEE International Conference on Robotics and Automation (ICRA)
Patents
System and Method for Guiding LiDAR-based 3D Object Detection by Multi-resolution Features Recovery Using Panoptic Segmentation Information
Yixuan Xu, Hamidreza Fazlali, Bingbing Liu
[patent filing since 2022]
Projects
deMISTIfy: Machine Intelligence Newsletters
A monthly machine learning newsletter, ongoing since 2021. Serving as the head editor.
Past Issues | Medium Articles | Subscribe!
E-mail newsletters featuring summaries of articles written by UTMIST's technical writers and me, covering interesting machine learning papers, news, and resources.
SketchToPicture: Paired Image-to-Image Translation using cGAN Conditioned on Class Prediction
A project completed with Sumin Lee and Jiarui Zhang in 2021. Served as the team lead.
A generative model that uses sketches or edge maps as input, classifies the depicted objects, and converts the sketch into realistic images. Used pretrained ResNet50 as classifier to provide labels for a label-supervised pix-2-pix that leverages conditional instance normalization
COVID-19 Detection Diagnosis and Segmentation Tool
A project co-developed with Jiarui Zhang, Tianshu Kuai, Yihan Ni, and Yunni Tang in 2020.
An encoder-decoder model for detecting COVID-like symptoms and segmenting affected areas on lungs given CT scans as a potential method for easier diagnosis during test kit shortages. Achieved 91% classification test accuracy given limited samples.
OpenRace: Bringing Runners Together in Virtual Races
A mobile app business stemmed from a startup accelerator, co-created with Benjamin Cheng, Yannis He, Stanley Zhang, and a team totaling 12 members in 2018. Served as the team lead.
Pitch Deck | App Demo | Promo Video
A mobile fitness app that enables runners to match with other similarly-skilled runners from around the world and compete in real-time races anywhere, anytime.