Shenghao XU (Singho HUI) 

Cloud Engineer

Center of Cyber Logistics
The Chinese University of Hong Kong
Shatin, N.T., Hong Kong

Email: shenghaoxu@cuhk.edu.hk


Biography

Shenghao XU is currently a cloud engineer in the Center of Cyber Logistics (CCL), The Chinese University of Hong Kong. Previously, He received his Master's degree in the Department of Computer Science and Engineering, The Chinese University of Hong Kong, in 2021, supervised by Prof. John C.S. Lui; and his B.Sc. (1st hon.) degree in 2020, under the supervision of Prof. Hung Kevin.

His research interests broadly lie in Computer Vision, Machine Learning, Deep Learning, and Online Learning. Narrowly in Self-Supervised Learning, Multimodal, and Multi-armed Bandits.

News

Publications

   
Development of an AI-based System for Automatic Detection and Recognition of Weapons in Surveillance Videos.
Shenghao XU, Kevin Hung
IEEE Symposium on Computer Applications & Industrial Electronics (ISCAIE), 2020.

[Paper][Code]

BanditMF: Multi-Armed Bandit Based Matrix Factorization Recommender System.
Shenghao XU, John C.S. Lui
arXiv

[Paper][Code]

Selected Projects

         
Literature Review: Deep Learning and Transformer.
This document, we explores advanced methodologies in deep learning for object detection and the transformative impact of the Transformer architecture in various domains.

[Report]

Decision Tree Classifier to Predict the Income of the Adult in America.
In this project, we proposed a modified version of Hundt's algorithm to construct a decision tree for predict the income of the adult in America. To avoid unreliability and overfitting occur in the subtree, we proposed a parameter called 'small factor' which shows how small the subset is, and by comparing the accuracy of the model, we can split the data-set more appropriately.

[Report][Code]

Ncnn-YOLOv3 Acceleration and Implementation.
This project designs and implements the porting of YOLOv3 to mobile and uses YOLOv3 for object detection. During the migration, NCNN, which is the high-performance neural network inference computing framework, is used to quantify and reduce the size of the YOLOv3 model, ultimately enabling acceleration without compromising detection accuracy on the mobile side.

[Report][Code]

Searchable Encryption.
In recent years, with the rise of GPU and cloud computing, the security of traditional cryptography has been challenged, which leads to the emergence of new cryptography technologies. In this report, the state-of-the-art cryptography technology-searchable encryption(SE) will be illustrated and discussed.

[Report]

Honors & Awards

Professional Activities

Experience

Teaching

2023-2024Fall Advanced Information Systems/Business Analytic Practicum (DSME 6686IS/6696BAP1)
2021-2022Fall Advanced Business Analytic Practicum (DSME 6696BAP1)

Miscellaneous



© Runze S.H XU | Last updated: Feb. 20th 2024