The Image Understanding Architecture (IUA) project was dedicated to building a specialized heterogeneous parallel architecture for computer vision research. The IUA consisted of three processing layers interleaved with two shared memory layers. At the lowest is level is the CAAPP, a bit-serial mesh-connected SIMD processing array. The CAAPP also incorporates a coterie network that supports associative processing through switchable connections between neighboring PE's. The ICAP is a layer of MIMD processors that communicate with the CAAPP through shared memory and with each other via a message passing communication network.
Three organizations participated in this effort. Amerinex A.I. was responsible for creating the IUA system software. The Visions group of the Computer and Information Science department at the University of Massachusetts/Amherst originated the concept of the IUA. Hughes Research Laboratory was responsible for the design and fabrication of the IUA hardware.
This work ws funded by the Army Research Laboratory and the Advanced Research Projects Agency as part of the UGV project under project numbers DAAL02-91-K-0047, DAAL02-91-C-0048, and DAAL02-91-C-0051. This work was continued by the Image Understanding Architecture Vision Computer Commercialization Consortium as part of the Technology Reinvestment Project (TRP) under agreement #MDA 972-94-3-0048.