For years, computer vision algorithms have been developed primarily on expensive workstations, using specialized and expensive peripherals (e.g. framegrabbers, cameras). Recent developments in PC hardware and operating systems have suddenly made suitable hardware available at much lower cost, and in a much smaller package. Specifically,
Our Computer Vision Handbook contains more information on specific hardware.
Development of new hardware is proceeding very quickly. The PC market is very large, much larger than the workstation or Mac market. Therefore, PC products tend to be smaller, more reliable, and less expensive. Standard home PC configurations now include multimedia hardware, such as color displays, graphics accelerators, cameras and scanners. The newer desktop PC's are approaching the speed and memory of workstations; the newer laptops are not far behind.
In an independent development, the Linux operating system and its cousin FreeBSD have become reliable. These operating systems are, effectively, implementations of Unix for PC hardware. It has become fairly straightforward to install them on standard desktop PC systems. Installation on laptops is still somewhat difficult, and support for certain peripherals erratic. However, its popularity is growing fast.
In particular, large numbers of computer scientists, both graduate students and faculty members, are buying home PC's and installing some version of Unix (typically Linux) on them. Some are still using Unix workstations for their research and teaching; some have switch to PC's running Linux.
We use HP's enhanced port replicator ($300) to provide a LAN connection and a SCSI-2 port. On the SCSI-2 port, we can attach an Iomega Zip Drive ($220) for image and file storage or a Panasonic CD-ROM player ($320). We also own a Snappy framegrabber ($200) and a Toshiba miniature color camera ($330), as well as an Epson PhotoPC ($500) for image acquisition, but these are not yet working under Linux.
For details on how this equipment works, how we installed Linux, and the like, see our Omnibook page.
This page is maintained by Margaret Fleck. and Dan Stevenson.