visibility-based pursuit evasion
We discuss a formulation of the visibility-based pursuit evasion problem for a class of "2.5"-dimensional environments. We specify the class of environments under consideration and the specifics of the pursuer and evader agents. We consider two kinds of pursuers: one with omnidirectional visibility and one with a limited field-of-view, and attempt to derive algorithms for solving the problem by extending existing algorithms for corresponding 2D cases. Finally, we analyze limitations of these approaches (and the class of 2.5D environments under consideration) and discuss potential avenues for future work.

Report

feature tracking and factorization from monocular video
Determining shape and structure of moving objects from a single-camera video of their motion has been an active research area. Good, robust solutions to this problem enable, for example, the detection of hand gestures and facial expressions of a human subject by observation from a single camera. The aim of this project is to study how feature tracking and motion factorization can be coupled to implement a complete, robust motion capture system that requires monocular input. We examine the popular KLT tracker, and some ways of making it more robust; and see how it works in conjunction with the Tomasi-Kanade factorization algorithm for rigid body motion.

Report Presentation

level of detail in computer graphics
As data sets used in interactive visualization applications grow in size and complexity, techniques for adapting the detail of objects in the scene to ensure interactive frame rates gain increasing importance. This report provides an overview of the general approaches used in such Level of Detail frameworks, and focuses on the problem of mesh simplification (since triangle meshes are currently the dominant type of model in most scenarios requiring LOD techniques). We examine a popular scheme called Progressive Meshes (PMs), and explore how the PM representation supports view-independent and view-dependent refinement during interactive rendering.

Report Presentation

quantum cryptography
A survey of some of the latest developments in the field of Quantum Cryptography. Among the topics covered are quantum computing technology and the challenges it faces, quantum communication channels and key exchange protocols, and quantum algorithms to crack existing cryptosystems (such as Shor's factoring algorithm).

Report

gcc tricore port
A study of the structure of GCC (with emphasis on Machine Descriptions) and the port of GCC to Tricore processor. We considered the difficulty of upgrading the port to GCC 4.0.0. We also focused on the task of cleaning up the existing port so that most changes are localized in the machine-specific directory of the source tree.

Report

face-off: a face detection and recognition tool
Face-Off is an easy-to-use face detection/comparison/storage tool. This is a multi-component system written in multiple languages, including C++, Java and Unix shell script. It uses several open-source libraries for backend processing. The basic algorithm used is the Eigenface technique.

Source