Page 85 - Undergraduate Research Opportunities Program Proceedings 2017-18

School of Engineering
Department of Computer Science & Engineering

polynomial time approximation scheme for this discrete version which can be used to derive a polynomial time
approximation scheme for the original version.

Supervisor: WONG Raymond Chi Wing / CSE
Student: SHI Zhenmei / COSC Course: UROP2100, Fall
UROP3100, Summer

This report is a progress report. Our project mainly focuses on designing and implementing a new visual of tracking
algorithm by using the trajectory method and deep learning method. Our goal is to participate in VOT 2018 competition
held by ECCV 2018. We reviewed the latest method in visual of tracking and proposed our own algorithm. The proposed
algorithm is still under construction, which may include deep reinforcement learning in action space, memory
augmented structure and a self-located mechanism like a grid cell. These ideas are coming from the latest development
in deep learning, tracking algorithm and neural science. We are now building the algorithm by using pytorch and training
our model on VOT and OTB datasets with GPU.

Supervisor: WONG Raymond Chi Wing / CSE
Student: ZHANG Jintao / SSCI Course: UROP1000, Summer

In this project, our main goal is to build a e-learning platform which provides course pathway planner and grade
prediction for students in HKUST using data mining. I am in charge of the web based user-interface design and the
integration of data-mining and pathway planner part. The web application is structured using Django, a python based
web-app framework, together with a Postgres database. The students demographic information and course history is
acquired using copy and paste from HKUST Student Center. All the students’ demographic information is stored in an
object with the private key of the student’s HKUST ID. And with a count model, the number of each course that is
inquired in the grade prediction part can be tracked and recorded, intended for further analysis.

Supervisor: WONG Raymond Chi Wing / CSE
Student: ZHOU Yeju / COMP Course: UROP1100, Fall

Terrain-Toolkit is a cross-platform tool which is workable on Windows, MacOS and Linux. The toolkit was developed for
visualizing and processing terrain data, allowing users to conduct several operations on data with different data formats
through a graphical interface. The toolkit p Algorithms and Games in Android Devicesrovides multiple operations in a
sole software, including surface distance calculation, surface simplification and format conversion. Nevertheless, the
toolkit was first implemented on Mac OS, and then extended to Windows and Linux, making the installation on Windows
tedious. A guideline of installation on Windows is provided in the package. As the toolkit was tested on Windows 7 a few
years ago, the obsolete installation guideline needs updating to achieve a smooth installation.

Efficient Queries over Database

Supervisor: WONG Raymond Chi Wing / CSE
Student: LIANG Geyu / MATH-IRE Course: UROP1100, Fall

In this final report, the work of me doing the UROP 1100 project on fall semster 2017 will be introduced. The ultimate
goal is to produce a compression algorithm that can be applied to the current format for the maps containing lane
information. In this report, first we’ll focus on the lane infromation in a map and its characteristics and advantages
campared with traditional map. Then we’ll do a simple review for a trajectory simplification method introduced by

75