University of Tokyo "page flip" scanner video

[Turtle]

The Japanese are investing on their students. But this is about the same trick used by Google filed on Sept 2004, but it could have been conceived as far back as 2002. Those who work at Google's scan bureau aren't allowed to disclose how fast they scan but 200 pages or more per minute sounds about there. The difference is that Google can afford processing hardware, high speed, high resolution cam, we don't. Some of us here built scanner from nearly zero in cash with free processors. That's a good thing, hardships spark creativity. We just have to wait until HD camcorder can shoot at around 6 or 7 megapixel at 500 frames per second for $300US.

Google_march-09.jpg (9.37 KiB) Viewed 6267 times

[daniel_reetz]

There are other possibilities, like pooling our resources, building a king-hell book scanner, and shipping it around to digitize all our libraries. I've often thought that we, as a community, could pool our workshops and make all the parts for scanners super-cheap, and sell them to each other. I would buy or build a CNC machine to get that done.

Or other, smaller solutions. I have two in mind, one in prototype. I'll share next month when this months' commitments are over.

[koulevprime]

There are other possibilities, like pooling our resources, building a king-hell book scanner, and shipping it around to digitize all our libraries. I've often thought that we, as a community, could pool our workshops and make all the parts for scanners super-cheap, and sell them to each other.

Speaking of which, some individuals from my University are investing in a Hackerspace, dedicated to the community as a workspace for people to hack together whatever they want. I will definitely be using it for working on my scanner!

[Afish]

Flipping the pages in that fashion will always result in missed pages. Looking for them and rescanning individual pages will make one wish he could do it right the first time.

[benjamin]

My understanding is Decapod is the low-cost, DIY equivlent of this... obviously it's different approach, but they're looking to interpolate data from multiple shots of the same page (in their case, one shot each from two cameras rather than 500fps from one).

I wonder whether the Tokyo system actually gets better results than ones like those in this community- he says the camera shoots 1000 x 1000 pixels, I'd like to know whether it straightens out a single reference image based on the others; or adds info from many shots to get higher rez.

I agree, however, that page flipping is still the weak link. 200 pages in 1 minute isn't actually all that much faster than our system (in theory we can do up to around 90 per minute). Until we can build robots to replace our clumsy ape hands in the process, the capacity demonstrated here is largely going to waste.