Projects

Air Rope Jumping APP

Dec 2015 – Apr 2017

  • Developed a rope jumping counter APP on Apple Watch that helps users to count the number of jumps and trips
  • Developed a virtual rope jumping APP on Android that enables users to do rope jumping with a cellphone. The app simulates the sound of swinging rope synchronously to improve sense of reality, and also count for users. Gained 10,000+ downloads on 360 APP Market
  • Tested machine learning methods to distinguish rope jumping from other motion and attained 90% accuracy, and built a demo to prevent users from fabricating data by doing other motion. Was recognized as National Undergraduate Training Program for Innovation, and received 15,000 RMB research funding by the Ministry of Education of the People’s Republic of China

 

Read More

Machine-Learning Based Toxin Classification

Research Assistant to Prof. Yau Shu Wong
Jul 2016 – Sep 2016

  • Used bioinformatics toxin classification to reduce costs for pharmaceuticals by replacing complex wet lab biological experiments; investigated machine learning methods based on the change curve of cell number in toxicant solution
  • Introduced a shapelet transformation for feature extraction and random forest for classification, and attained 94% classification accuracy, a 5% improvement from state-of-the-art papers; further improved shapelet transformation by introducing discretization of attributes, and raised the accuracy by another 3%
  • Presented research findings at Banff Venture Forum 2016, Canada’s premier company financing event

Read more

Intelligent Irrigation Control System Design

Research Assistant to Prof. Yau Shu Wong
Jul 2016 – Sep 2016

  • Submitted a paper to Catena as a co-author
  • Assisted a startup in designing a water-saving irrigation sprinkler by providing mathematical models and control algorithms that could counter the effects of wind, avoid out-of-bounds spraying, and attain uniform water distribution
  • Individually implemented the algorithms in C; assisted a colleague to deploy it on the Arduino control module; Optimized the control algorithms, improving the efficiency by 7.3 times and the precision by 2.5 times
  • Submitted a paper to Catena as a co-author

Read more

VAST Challenge Visualization

Dec 2015 – Jan 2016

  • An solution to IEEE VIS VAST Challenge, using the GPS data of all the visitors in the theme park to find out a thief
  • Preprocess GPS data into practical features, used PCA to reduce dimentionality, and used K-Means to classify and find out the abnormal visitors.
  • Built heat map to analyse the whole traffic of the park, and build path tracker to track the abnormal visitors to find out the suspect.

Read more

E-Business Sales Data Correlation Analyses & Product Bundling Suggestion System

Apr 2015 – Jun 2016

  • Constructed crawler program to collect sales data from Alibaba platform; used d3 to visualize sales data by a self-designed dynamic interactive streamgraph and tree map to display correlation among data
  • Introduced Granger causal relation testing to analyze association and correlation factors for tracking among different products in online stores, and to generate suggestions on how to bundle products to boost sales

 

MiniSQL

  • A self-implemented database system like MySQL
  • Implemented B+ tree index, clock buffer management algorithms, query parser and so on

Read more

Image Processing

  • Implemented a image processing software using MFC framwork
  • Implemented bilateral filtering, Laplacian enhancement, histogram equalization and lots of other image processing algorithms

Read more