The CAMEO system consists of a set of four or five firewire cameras arranged in such a way to capture a full panoramic view of the world. These cameras are interfaced into a small form factor (shuttle) PC so that the images can be turned into a consistent mosaic. The PC is and the camera head are shown below. Once captured, the five raw video data streams are merged into a consistent panoramic image mosaic. This mosaiced image is passed into the face detection module that returns the (x,y) (image coordinates) positions of all of the faces in the image. This data is broadcast to other computers over TCP sockets and is also saved to a log file. Other machines running a GUI can observe the data returned from CAMEO in real time as well as replay the log files offline. Since CAMEO is designed to link into the larger CALO system, it would have access to a knowledge base that would provide information regarding the meeting context. In the figure, these connections are denoted by dashed lines. A conceptual diagram of CAMEO is shown below. The individual images are merged into a single consistent mosaic by identifying shared feature points in each of the adjacent images. Once merged, the images are searched for faces using the Schneiderman/Kanade face detection algorithm: H. Schneiderman. "Feature-Centric Evaluation for Cascaded Object Detection." IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2004. H. Schneiderman. "Learning an Restricted Bayesian Network for Object Detection." IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2004. The following image shows an example image with the identified faces highlighted.