Recent needs for high performance computer vision techniques led to using GPUs as a powerful processing platform for various computer vision algorithms. This development was driven by the ability of many computer vision algorithms to use standard computer graphics techniques to their benefit, the high performance gains of GPUs in recent years and their increasing ability with to support general purpose computing. These trends gain more and more speed and the workshop brings computer vision researchers together that want to participate in this exciting development of fast computer vision algorithms. In the last years we have already seen enormous progress in this area that led for example to real-time stereo techniques but there are still many challenges left for a variety of other problem classes. The workshop invites contributions that tackle problems of computer vision employing the unique capabilities of GPUs to improve the performance or propose new methods targeted to GPUs. Lunch Tutorial: “GPU programming in CUDA”, Joe Stam, Nvidia "Over the past several years many researchers have explored the use of GPUs for accelerating computer vision applications. GPU architectures have also evolved over this time to better support general purpose high performance parallel computing. NVIDIA's latest GPU architecture fully supports a wide variety of general purpose applications through CUDA, a new 'C' based development platform for GPU computing. "This tutorial session will introduce CUDA and NVIDIA's new GPU architecture in the context of programming computer vision applications. We will present the CUDA programming model, walk through some introductory examples, and demonstrate high-performance CUDA implementations of some common vision algorithms." The lunch for tutorial attendants will be provided.

Workshop program

Recent needs for high performance computer vision techniques led to using GPUs as a powerful processing platform for various computer vision algorithms. This development was driven by the ability of many computer vision algorithms to use standard computer graphics techniques to their benefit, the high performance gains of GPUs in recent years and their increasing ability with to support general purpose computing. These trends gain more and more speed and the workshop brings computer vision researchers together that want to participate in this exciting development of fast computer vision algorithms.

In the last years we have already seen enormous progress in this area that led for example to real-time stereo techniques but there are still many challenges left for a variety of other problem classes. The workshop invites contributions that tackle problems of computer vision employing the unique capabilities of GPUs to improve the performance or propose new methods targeted to GPUs.

Lunch Tutorial: “GPU programming in CUDA”, Joe Stam, Nvidia

"Over the past several years many researchers have explored the use of GPUs for accelerating computer vision applications. GPU architectures have also evolved over this time to better support general purpose high performance parallel computing. NVIDIA's latest GPU architecture fully supports a wide variety of general purpose applications through CUDA, a new 'C' based development platform for GPU computing. "This tutorial session will introduce CUDA and NVIDIA's new GPU architecture in the context of programming computer vision applications. We will present the CUDA programming model, walk through some introductory examples, and demonstrate high-performance CUDA implementations of some common vision algorithms." The lunch for tutorial attendants will be provided.