Four MSc theses to be presented on October 19, 2017 and one extra on October 23, 2017.
October 19 is the day for coordinated master thesis presentations in Computer Science at Lund University, Faculty of Engineering. Four MSc theses will be presented.
Note to potential opponents: Register as opponent to the presentation of your choice by sending an email to the examiner for that presentation (firstname.lastname@example.org). Do not forget to specify the presentation you register for! Note that the number of opponents may be limited (often to two), so you might be forced to choose another presentation if you register too late. Registrations are individual, just as the oppositions are! More instructions are found on this page.
|TITLE||Testing how a object relational mapping library is mapping data between SQL and Java types|
|SUPERVISORS||Emelie Engström (LTH), Per Minborg (Speedment AB)|
This thesis presents the Mapper Testing Tool, a program aimed to test how the ORM is mapping data between SQL and Java types, focusing on improving the efficiency on develop, execute and evaluate tests. To increase the efficiency the Mapper Testing Tool test execution is semi-automated and the test cases can be reused between databases.
|TITLE||Adaption of a new IPC on the Linux Operating System|
|SUPERVISORS||Jonas Skeppstedt (LTH), Umut Tezduyar Lindskog (Axis Communications AB)|
Interprocess communication (IPC) is an essential part of any operating system. In order to fully function, programs must be able to communicate with each other in some way, be it directly through a message stream or shared memory, or indirectly through writing and reading files between one another. This thesis looks at a new IPC primitive, Bus1, comparing it to existing alternatives on Linux, namely Unix Domain Sockets and D-Bus. This is achieved through a proof-of-concept implementation of Bus1 on an Axis camera, replacing parts of the current D-Bus communication with that of Bus1, and is evaluated by comparing transmission times of messages. Since small devices, such as Axis cameras are often resource limited in terms of size, software efficiency becomes very important to maximize performance.
|TITLE||General Methods for the Generation of Seamless Procedural Cities|
|SUPERVISOR||Michael Doggett (LTH)|
|ABSTRACT||Procedural generation as a concept is as old as computer graphics. It is usually defined as a method for creating data algorithmically as opposed to manually. Work in this area often revolves around noise functions and large degrees of randomness, which works well for chaotic structures and for generating natural environments, but it does not always suit more sophisticated and coherent environments. This thesis proposes some new approaches for procedural generation, specifically on the subject of cities. Many of the solutions discussed are applicable in other areas of procedural generation where order and coherency is important as well. Work on procedural cities has been done before, but the scope has usually been limited to one or a few aspects of the city. This project combines several of these previously explored areas and several new ones into a composite whole.|
|TITLE||Minimising Light Leakage in Real-Time Photon Splatting|
|SUPERVISOR||Michael Doggett (LTH)|
|ABSTRACT||Ray tracing from pixels to splatted photon positions is an efficient method of reducing light leakage from photons with large radii. In this thesis we show how we determine leaked photon energy for culling by shooting a ray from the pixel position to the position of a contributing photon and detecting hits from geometry intersection. We show the level of leakage can be significantly reduced, resulting in greatly improved lighting realism.|
|Monday October 23|
13:15 (room E:2116)
|TITLE||Shopping list generation with Machine Learning|
|SUPERVISOR||Pierre Nugues (LTH)|
|ABSTRACT||This master’s thesis attempts to use data from grocery receipts and predict future receipts. The items or groceries on the receipts are categories into two layers of categories, 209 sub categories and 17 main categories. These categories are model as time series with an indicator variable that used to model purchase/no purchase in the category. This indicator variable is estimated by using Linear Support Machines in combination with an intensity expectation. While the amount is modeled as a Gaussian field and estimated with ordinary kriging. The provided data contains 15969 data points, from 1230 receipts and 34 household. The amount on the item is measured by using the price paid for the item.|