lu.se

Datavetenskap

Lunds Tekniska Högskola

Denna sida på svenska This page in English

Seven Computer Science Master Thesis Seminars

2014-01-17

January 17th was the day for the second coordinated master thesis presentations in Computer Science at Lund University, Faculty of engineering. The plan is to arrange these kinds of presentations 4-5 times a year. The event is open to all interested in listening to one or several seminars, as desired.

When: Friday January 17 at 9-16

Where: E:2116, E-huset, Ole Römers väg 3, Lund University

09.15 Edvin Berglin: On the performance of edge coloring algorithms for cubic graphs

10.15 Therese Alenlöv: Off-line Time Synchronization in a Log Analysis Tool

11.15 Rasmus Persson: Motion blurred real-time ray tracing

12.15 Martin Jacobsson: Progressive Photon Mapping using Cloud Computing

13.15 Alexander Neckmar: Detecting Unique Patterns in Human Head Rotation via Low Resolution Camera Sensor and Imaging Software

14.15 Robin Palmblad & Joachim Nelson: Video conference communication on AXIS cameras

15.15 Daniel Raniz Raneland: Performance- and Cost-efficient Cloud Architectures

Details below or at: http://cs.lth.se/examensarbete

Recommended reads from the first occasion:

kl 09.15: On the performance of edge coloring algorithms for cubic graphs

Författare/Authors: Edvin Berglin

Handledare/Supervisor: Thore Husfeldt (LTH)

Examinator/Examiner: Jonas Skeppstedt (LTH)

Abstract: 
This thesis visits the forefront of algorithmic research on optimal edge coloring of cubic graphs. We select a set of algorithms that are among the asymptotically fastest known today. Each algorithm has exponential time complexity, owing to the NP-completeness of edge coloring, but their space complexities differ greatly. They are implemented in a popular high-level programming language to compare their performance on a set of real instances. We also explore ways to parallelize each of the algorithms and discuss what benefits and detriments those implementations hold.

kl 10.15: O-line Time Synchronization in a Log Analysis Tool

Författare/Authors: Therese Alenlöv

Handledare/Supervisor: Jörgen Bergström (SAAB)

Examinator/Examiner: Pierre Nugues (LTH)

Abstract: 
The work of analysing log files has been facilitated by the program Logan, developed at SAAB AB. However, Logan only had support for viewing one log file at the time, wherefore a desired upgrade was requested. In this thesis, the making and implementation of a plug-in to Logan is reported, which can handle multiple log files and is user friendly. A problem that is raised when viewing multiple files is time synchronisation, since the events of the log files where time stamped by different clocks. It is therefore proposed a linear model for every clock, which is used to find a global time stamp for every event. The parameters of the model is found by the use of global events, events that are logged in more than one log file, and by using linear regression on these global events, we can find a relation between the different local clocks.


kl 11.15: Motion blurred real-time ray tracing

Författare/Authors: Rasmus Persson

Handledare/Supervisor: Michael Doggett (LTH)

Examinator/Examiner: Tomas Akenine-Möller (LTH)

Abstract: 
Computer graphics become increasingly important in today’s technological society and demands on performance and photo realism in 3D-applications are rising. Rasterized graphics are usually used in real-time graphics but ray tracing offers better accuracy for shadows, reflections and refractions, thereby improving photorealism. The purpose of this study is to find a way of accelerating motion blurred ray tracing. This thesis focuses on motion blur in real-time ray tracers and presents a BVH-building strategy to increase tracing performance for moving triangles. By dividing the scene not only in space, but also in time, unreasonable large bounding boxes can be avoided. The result from the study shows improvements in performance for scenes with much movement. These results show that ray tracers and accelerating data structures can be optimized for better real-time performance. Adopting tracers and accelerating data structures for real-time rendering will improve performance, driving the development towards the ultimate goal: ray traced photorealistic 3D-games.


kl 12.15: Progressive Photon Mapping using Cloud Computing

Författare/Authors: Martin Jacobsson

Handledare/Supervisor: Rasmus Barringer (LTH)

Examinator/Examiner: Tomas Akenine-Möller (LTH)

Abstract: 
Progressive Photon Mapping is a robust global illumination method capable of computing unbiased and consistent solutions. Using a limited amount of memory it is often effective in producing a high quality result for scenes involving complex light transport, such as reflections of caustics. However, response times are limited by the executing computing platform.The rise of cloud computing platforms have provided us with a cost-efficient way to employ advanced light simulations on computers available for rent by the hour. This thesis introduces a variant of Progressive photon mapping suitable for parallel execution in a cloud computing environment and a framework for deploying this on the Amazon EC2 platform.


kl 13.15: Detecting Unique Patterns in Human Head Rotation via Low Resolution Camera Sensor and Imaging Software

Note: bachelor of science/kandidatarbete (15hp)

Författare/Authors: Alexander Neckmar

Handledare/Supervisor: Paul Cronholm (Crunchfish AB)

Examinator/Examiner: Elin Anna Topp (LTH)

Abstract: 
This paper is a study of ways to recognize horizontal rotational head movement using existing software capable of pinpointing face position in images from video streams. The software makes use of the embedded camera in laptops and mobile devices. The study makes use of the following parameters: - The center of the face’s horizontal (x) and vertical (y) position. - The width of the face. This paper describes three methods for making this distinction.

 

kl 14.15: Video conference communication on AXIS cameras

Författare/Authors: Robin Palmblad & Joachim Nelson

Handledare/Supervisor: Jesper Olavi (AXIS)

Examinator/Examiner: Mathias Haage (LTH)

Abstract: 
Video conference is a growing market. Wouldn't it be great to be able to make a video conference call only using cameras? That is the focus of this thesis. To be able to do this we have implemented the functionality to receive a video stream over the network and send it out on a camera's video output.

A network connection varies over time and adaptive bitrate streaming is needed. A stream’s QoS is evaluated to decide how the rate controller will change the video quality to maintain low delay and still keep a good QoE. Different algorithms are tested in order to come up with the most suitable option for our system.

The algorithms performs well in adapting to the network load. However, when packages are lost the GOP will be disrupted and the video quality impared. Some improvements can possibly fix this and by using TCP instead of UDP.


kl 15.15: Performance- and Cost-efficient Cloud Architectures

Författare/Authors: Daniel Raniz Raneland

Handledare/Supervisor: Vadim Feldman (Flygprestanda AB)

Examinator/Examiner: Mathias Haage (LTH)

Abstract: 
The Cloud is becoming a more and more popular expression and as companies learn what it is and the cloud services offered by cloud providers mature, more companies are looking towards The Cloud to satisfy their needs. This master thesis examines issues that can arise when moving an existing server solution to The Cloud and how to utilise the possibilities offered by cloud services with a cost-effective solution. This is achieved by developing a cloud-deployment architecture for an existing server solution, FOCS, for deployment on Amazon’s EC2 service and then migrating FOCS to this architecture. The deployment is then optimised on both performance and cost-effectiveness - both on a source code level through refactorisation of any bottlenecks that arise and on an architectural level by modifying the architecture to avoid performance bottlenecks associated with network throughput.



Exjob@cs


Till lista.
Sidansvarig: