Maryland CPU-GPU Cluster Infrastructure

Real-Time Computer Vision

Our work in real-time computer vision applications is based on distributed cameras. By using distributed cameras, traditional problems confronted when using a single camera such as occlusion, disappearance and reappearance of objects, and recovery of 3D motion trajectories of small objects, can be conveniently addressed. The specific problems we plan to address on the CPU-GPU cluster are: detection, tracking and fusion of trajectories using distributed cameras, view synthesis using image based visual hulls, gait-based human recognition and human activity analysis.

Human Tracking at a Distance

We plan to experiment with an outdoor distributed camera test-bed of eight cameras that have been self-calibrated. A CPU-GPU cluster provides a powerful engine to do real-time image processing for multiple video feeds as well as to visualize the results from novel views that are not represented by the distributed cameras. We plan to deliver the video feeds to the Infiniband network which will then send the video data into the individual CPU-GPU workstations.

Human Activity Analysis