I've built a book scanner from that site a number of years ago and it has worked well enough for what I needed.
The real problem isn't the hardware though, it's the multiple programs needed to process the images and get everything into a small text searchable pdf file afterwards.
To give you an idea, my workflow usually starts by importing all the left pages into Lightroom, process for things like correcting blacks and whites, keystoning, skewing, and cropping, and then I export everything as jpg files. I then repeat for the right pages. This has to be done separately because each camera sees the book from a different angle -- lighting is usually different, keystoning will be different, and even the distance of the camera to the page has to be taken into account for correct cropping. After that, I run a perl script to combine the left and right pages so that they're numbered sequentially, and then finally import into Adobe Acrobat Pro to make text searchable pdf files. I've tried all other OCR software, and Acrobat has them all beat. If there is color in the images the pdf file will be HUGE. I've scanned some of my son's books for school that were in color and attempting to view them on an iPad 3 was folly.
There is a program called Scan Tailor that also helps process images. It does a decent job of finding the borders of the pages for auto-cropping, and attempts to correct skewed pages, however it requires looking through each page to make sure it's found everything correctly. Too often I'll find it crops incorrectly, missing things like page numbers in the corners of the page.
When I'm looking to make the smallest PDF files possible, I'll use this after Lightroom.
This indigogo campaign looks to make this whole thing a lot simpler (Czur Scanner):
It's apparently an all-in-one solution with hardware and software. The video shows it doing black and white well enough, but I question how well it will deal with color (They don't show any demos of color books). Seemed good enough to purchase (I did so), even if only for black and white, and simpler than the DYI setup I've been using.