Processing 3D Geometry for Modern Graphics Hardware

Budirijanto Purnomo, Johns Hopkins University

Recent advances in graphics hardware technology have transformed Graphics Processing Units (GPUs) as powerful parallel machines. Modern GPUs consist of hundreds of processors with high performance memory. In order to take advantage of these hardwares, we need to design data structure and algorithms that adapt well to this architecture. Due to the advances in 3D scanning technology, increasingly detailed 3D models are used in movies, games or simulation. These models range from millions to billions of triangles. In this talk, I will show how to process these 3D models for efficient processing by modern GPUs. I will present seamless texture atlas (STA) as a new GPU-friendly efficient representation for 3D geometry suitable for large data processing. Using STA, we push state-of-the-art in three graphics applications: seamless texturing of large 3D models, view-dependent visualization of huge models and geometry processing of 3D models.

Speaker Biography

Budirijanto Purnomo was born in Pontianak, Indonesia. After completing his high school in Jakarta, Indonesia, he came to United States for advanced study. He obtained BS and MS in Computer Science from Michigan Tech University in 1999 and 2001. Then, he pursued a Ph.D. in CS with specialization in computer graphics at Johns Hopkins University. During his study, he completed two internships with Lawrence Livermore National Laboratory and Pixar Animation Studios. After graduation, he will join 3D Application Research Group at AMD. His research interests are interactive computer graphics, visualization and compiler optimization. Read more