Grant: Research Tools and Instruments - Category 1 (<$150,000)

We request 10 ""Big Performance"" workstations (T7610) from Dell, including up-to 28-core CPUs, up-to 512 GB RAM, and massively parallel GPUs and GPGPUs for scientific computing and visualization. These scalable workstations are essential for developing and testing fast combinatorial optimization methods for multi-dimensional big image data analysis in computer vision and biomedical applications. Our joint laboratory is internationally recognized for highly-innovative graph theoretic algorithms in image analysis. These methods are now considered standard in industry (GE, Google, Microsoft, Siemens, Adobe) and in academia all over the world. Many of the image analysis problems of 7-8 years ago are now routinely solved via methods developed in our lab. Our publicly available code is widely downloaded and licensed. However, new challenges arise when applying such methods to large-scale video and big biomedical data of ever growing complexity, size, and dimensionality. Recently widespread availability and improved quality of consumer cameras and medical imaging technologies (e.g. CT, MRI, micro-robots) now produce terabytes of exceedingly high-resolution 3D and 4D images or video. Efficient and accurate analysis of such big image data is critical for a large number of applications concerning safety, security, personal entertainment, and health. For example, new levels of efficiency and accuracy are needed for large-scale multi-view reconstruction (e.g. Google Maps), autonomous navigation (cars, medical robots), and computer-assisted diagnosis of major diseases (spinal cord, cardiac, retina, etc) affecting millions of patients in Canada.We expect to develop novel mathematical models, design and test innovative algorithms scalable to big image data. Following well established traditions of our lab, we will make software libraries publicly available for use at Canadian and international institutions.