Thesis/Project Final Exam Schedule

 

Final Examination Schedule

PLEASE JOIN US AS THE FOLLOWING CANDIDATES PRESENT THEIR CULMINATING WORK.

Autumn 2018
 

Thursday, November 29

Ge Zheng

Chair: Dr. Yang Peng
Candidate: Master of Science in Computer Science & Software Engineering
11:00 A.M.; DISC 464
GlobalFlow: A Cross-Region Workflow Service for Orchestrating AWS Serverless Computing Services

Cloud computing is changing the world by enabling pervasive access to shared pools of computing, storage, network and other software resources. Serverless computing, as a unique service in modern cloud services, plays an increasingly important role in the whole picture of cloud computing. Major cloud service providers such as Amazon AWS and Microsoft Azure have released their serverless computing services and serverless orchestration services successively. In this project, a workflow service called GlobalFlow is designed and implemented to coordinate various dependent AWS Lambda functions residing in different regions. Such a cross-region orchestration service enables developers to select how to execute a series of Lambda functions given the limitation of data and/or computing resources. In GlobalFlow, two major orchestration strategies, namely the copy-based strategy and the connector-based strategy, are developed. Specifically, the copy-based strategy can copy Lambda functions from different AWS regions to one single region per user’s specification and execute them there. The connector-based strategy, on the other hand, creates various observing connectors in the region where a Lambda function resides without copying them over to different regions. The evaluation results show that the workflow service can effectively execute a sequential or parallel workflow consisting of Lambda functions in different AWS regions without introducing significant overhead.

Hao Wu

Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering
1:15 P.M.; DISC 464
Chinese Landmark & Attraction Recognition Computing Services

With the development of deep learning algorithm and computing hardware, image classification technology has been remarkably improved over the past years. To continue advancing the state-of-the-art in computer vision, many researchers are now putting more focus on fine-grained recognition problems. For example, instead of recognizing general entities such as buildings, lakes and mountains, researchers are developing machine learning algorithms capable of identifying Seattle Public Library, Lake Washington or Mount Rainer. Landmark recognition is a useful yet challenging task, due to the lack of large annotated datasets. Previous related work neglects non-English speaking area since they collect data using English only. The commercial landmark recognition APIs review also indicates the incompleteness of Chinese landmark data. This project releases Chinese landmark & attraction dataset for recognition of human-made and natural landmarks in China. The dataset contains 42,548 images depicting 987 unique landmarks in China. Moreover, we build a landmark recognition model with two Convolution Neural Network representations (Residual Network and VGG) and presents the results. Lastly, an iOS application is built based on our landmark recognition model. It allows user to do landmark prediction on images taken by camera and images from device gallery.

Friday, November 30

Joshua Lanman

Chair: Dr. Kelvin Sung
Candidate: Master of Science in Computer Science & Software Engineering
1:15 P.M.; DISC 464
Volumetric Fog

Advances in computing hardware and the introduction of the Graphics Processing Unit (GPU) have empowered computer graphics in many fields, including the movie and video game industries. These advances have led to the widespread use of complex and realistic special effects such as explosions, fire, smoke, rain, fog, dust, shafts of light or shadow, and others. While many public displays of these effects in use exist, the details of their illumination modeling and implementations are often protected as trade secrets. This is especially true in the video game industry, where the difference between being a leader or just another competitor in the large sea of companies and independents can come down to the small quality achievements one company’s product has over its competitors.

The focus of this project is the understanding and implementation of one volumetric effect, fog. Our project has three goals. The first goal is to build an understanding of the illumination model of fog in general and how this effect can be approximated and created in modern imagery using a GPU. The second goal is to demonstrate our understanding by implementing an interactive system where these effects can be interacted with and examined. Our third goal is to provide a starting point for future researchers and enthusiasts by sharing both the lessons we have learned and the resources we have developed during this project.

Many challenges exist in creating realistic volumetric fog. In general, the volumetric illumination by itself is a complex process to model. The challenge becomes even greater when we consider the interaction between other objects in the scene and the fog particles. Add to this the fact that real fog varies in density as a function of distance, altitude, and time. Faced with an overwhelming number of calculations, modern techniques seek to strike a balance between hardcore mathematical modeling, advanced hardware techniques, and approximation to meet quality and performance goals.

In this project, we studied both the science behind light interactions with a participating medium and different techniques for modeling this behavior, culminating in the system we have implemented. To meet our goal of providing a starting point for future researchers and game enthusiasts, we have chosen to implement our project in the free, cross-platform game engine, Unity. Our system can generate fog either scene-wide or within finely controlled fog volumes at specified locations within the scene. The developer can control the density of the fog as a function of both distance and altitude. The developer can also introduce random, variable density into their fog along with velocity settings to simulate wind effects. This system also supports fog interaction with multiple lights within the game scene, as well as shafts of light and shadow within the fog volume. Each effect can be independently switched on and manipulated, giving the user precision control over the resulting fog. Our system, with its well-documented implementation, is an excellent learning and exploration tool for anyone interested in volumetric fog.

Thursday, December 6

(Henry) Ryan Harasimowicz

Chair: Dr. Brent Lagesse
Candidate: Master of Science in Cyber Security Engineering
3:30 P.M.; DISC 464
ACRAS – A Hybrid Graphical User-Authentication System

The traditional text-based password is ubiquitous in today’s computing environment, yet creation and maintenance of both usable and secure passwords remains one of the largest challenges in modern computing. This project investigates an alternative authentication mechanism to the traditional static text-based password. The Algorithmic Challenge/Response Authentication System (ACRAS) is a single-factor, one-time-password system based on the accurate recognition and interpretation of user-defined characteristics within a set of challenge graphics. There is broad consensus that the human mind excels at graphic recognition and cued recall when compared to the abstract rote memorization of a complex string of text. ACRAS leverages this innate ability of the human mind; providing a framework for system users to define a set of rules for the recognition and processing of select characteristics of graphic challenges. Application of these easily-recalled rules deterministically generates a one-time-password string that is dependent upon the session’s randomly selected set of challenge graphics. As a one-time-password system, ACRAS is inherently resistant to some of the more common attacks on traditional authentication systems and demonstrates an increase of protection against others as compared to these static systems. A series of user-experiments have been conducted with an ACRAS prototype to gauge usability and overall user impression of the system. 

Friday, December 7

Swetha Tayalur

Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering
1:15 P.M.; DISC 464
SAFEU, A Campus Safety Mobile Application For University of Washington, Bothell

The aim of this project is to build a mobile campus safety application for the University of Washington. The goal is to address the lack of medium for university community members to easily report safety related incidents to authorities such as campus safety department (CSD) or police and reduce the communication gap between the two, using smart phone technologies.  The mobile application would contain the following features: 

(1) Messaging System, to let the authorities and loved ones know about Incident (2) Location based Push Notification System, to interact with fellow app users and authorities in the vicinity (3)    Individual Location Tracking, to track the last known location of the user (4) Audio and Video Recording, for evidence capture capability (5) Advertisement Model, to monetize the application and (6) Campus Safety Tutorial, to raise awareness among individuals.

Back to top

Questions: Please email cssgrad@uw.edu