It'sa me! Maario!
Projects
Current Work An all in one pedestrian detection, limb segmentation, gesture recognition and classification system as a joint project between myself and two students in the Multimedia and Networking Lab. I am responsible for finding a system which will detect pedestrians in each frame and output their approximate locations and limb positions.
Applying Parallel Design Techniques to Template Matching with GPUs To appear in Proceedings of IEEE CVPR 2010 (pdf)
Aribitrary Time Template Matching Manuscript submitted to Elsevier's CVIU (pdf)
Fixed Time Template Matching Published in Proceedings of IEEE SMC 2009 (pdf)
A Near Optimal Acceptance-Rejection Algorithm for Exact Cross-Correlation Search In Proceedings of ICCV 2009, accepted as a poster. (pdf of poster, manuscript)
A Dual Bound Algorithm for Very Fast and Exact Template-Matching An extended paper based on the work above. It has been accepted to IEEE PAMI with minor revisions. (pdf)
Distributed fault tolerant voting system (2008) This was done for the advanced operating systems class I took, but out of all the class projects I've done, I thought it was the most interesting. Code available on request.
Stackboost (2007) This was a project using the WEKA machine learning package which increased the rate at which Adaboost converged. Code available on request.
Sudoku puzzle solver and generator (2007) This was done for my AI course my first year as a Masters student at UTD, though I went back and improved it later to run as fast as the fastest C implementations (it is in Java). It does occasionally, 'hang' though! Code available on request.
Genepage (2006) This is system for creating point and click editable pages in PHP with arbitrary layout using css. In case whoever is reading this is wondering, this page is not written using Genepage because it would simply be overkill. However, I do plan to update it. Code available on request.
Evolving Soccer in Breve making use of PUSH genetic programming language (2006) This was a part of my senior thesis at Hampshire College on adding diploid genetics to genetic programming. The code, however, is large and messy as it involves scattered modification of both Breve and PUSH from the ground up, so it's very unwieldy. While I now know about diffs and patches, I didn't at the time.