Content-based multimedia information retrieval using GPGPU Programming

Background

“Fermi” is the most advanced GPU computing architecture created by Nvidia Labs as a supercomputing device. This architecture delivers many innovations like the Nvidia GigaThread engine and the parallel DataCache technology. The 512 CUDA cores accelerate the performance on a wider array of applications than ever before. Additionally, the Fermi chip possesses memory that is accessible to all GPU cores in a shared memory manner. In the context of previously implemented application for retrieving chosen objects from a multimedia Database, it is intended to build a system extension that is some kind of special benchmark for the purpose of analyzing the system efficiency. Through choosing different measures and constructing a new form of work distribution accordingly each time a different measure used, an evaluation of its efficiency should be obtained.

The main goal of this Bachelor thesis is to evaluate the performance of this system and to further optimize it with respect to different forms of database requests.

Assignment

• General Understanding of the conception of parallel programming
• Analysis pre-existing information retrieval system
• The implementation of a new Analyser to evaluate different methods for solving database retrieval problems

We offer

• Comfortable working environment
• A good equipped computer infrastructure
• Personal and engaged supervision

Profile of requirements

• Prerequisites for starting a Bachelor's thesis
• Good programming skills in C/C++
• Competence to work autonomous and creative

Contact