"Frankencamera" project at Stanford, open source digital cameras

[rxmd]

Here's an announcement from Stanford university:

"Stanford professor Marc Levoy plans to develop and manufacture the "Frankencamera" as a platform that will first be available at minimal cost to fellow computational photography researchers." (press release)

Basically it looks like they're developing an open source camera platform, where everything is programmable and can be swapped in or out. On the basis of that, basically everyone who is into tinkering can build their own hardware and software, with the difficult bits like sensor readout and processing, framework for hardware interaction and integration etc. already taken care of.

There's some videos that shows what they're doing.

It looks big and ugly at the moment, but then, if you look at this you'd have thought "this digital photography thing will never take off".

 

[rxmd]

(On a more personal note, even though the prototype has Canon lenses strapped in front of an N95 cell phone sensor, I find this much more interesting than the umpteenth announcement of leaked photoshopped M9 pictures somewhere.)

 

[Matus]

Finally - if one can write software for its PDA, why should we not have the possibility to write apps or even directly work on the software inside the camera. Well - one may screw up - but the you just upload the original software back again.

I really like this - I was already considering how great it would be to have a program inside a digital camera that would allow the camera to behave like an advanced light meter if necessary. Let's see what comes out of this.

 

[re_visible]

Well, what is the point of this?
I think they are way too late...

There are DSLR out there that have HDR in the cam (Pentax is one i think)

There is a whole community of Canon hackers, getting RAW and 1/8000s out of their cheap P&S since some years now...

How is this going to change anything in D-photography, that is so driven by hardware?

And all that hardware developement is fully in the hands of few corporations that have the bucks to pay for research and to take the risk of failure in the market?!

In the end, they will only be hacking some existing hardware and trying to get out the max on functionality out of it... so what's the big deal?

 

[rxmd]

This is about something different than hacking the firmware of your Canon P&S to re-enable features that Canon disabled for you. Basically it's about creating a framework for camera-related development.

All cameras today are made from components that are largely off-the-shelf. Companies integrate these components into packages. You're basically limited to what they offer you. If you're unhappy with that, you can only go the Huw Finney route and start to re-do everything from scratch. This project is about creating an architecture and a reference implementation of a working camera from a set of off-the-shelf components, where later you can go and change everything about the system in a modular fashion. Basically it aims to take the necessity for having big bucks out of academic and hobbyist camera-related projects.

Want to use a different sensor? Just design the glue logic and write a driver for reading out the new sensor, all the processing logic remains in place. Want to experiment with new image processing routines? Leave your camera as it is, change this particular piece of software, the source code is all available. Want to convert your color sensor into a B/W sensor with four times the pixel count? Take out the colour filter, change the processing routines, there you are. Want your camera to output JPEG2000 instead of plain JPEG? Go ahead. Want to build an adapter to use Canon EF lenses on micro-4/3 bodies? Take their reference implementation of the Canon lens protocol and write a little program for a microcontroller that translates between the two protocols - so he individual components are useful, too. Want your camera to support both Canon and Nikon lenses with autofocus and open-aperture metering? Reverse engineer the Nikon lens protocol, plug in a driver and you're all set. This is on a whole new level than hacking existing hardware or flipping some bits in stock Canon firmware.

This is mainly for hackers who aren't afraid of doing hardware and/or software development on their own, just like computers were in the late 1970s and early 1980s. However, for those it opens quite a number of possibilities.

 

[Frankie]

I concur with "rxmd"

The digital imaging world is far bigger than RFF and criticisms over Leica M8...

Check out "ADS-80"; "Z(I DMC or RMD"; or "Ultracam X" and find out the cost and users' plights before uttering ignorance.

I bet those folks having to pay upwards of $1.5 million for a butt-ugly, high-maintenance system form arrogant-as-hell vendors :bang:will be very happy.

To put it in a different way: any RFF member ever seen a 9" x 9" (23cm x 23cm...almost a square foot size) format aerial camera that uses 400' roll film with lenses the size of a bucket? Such cameras were cheap starting at $500k.

 

[pebcak]

Interesting... good to know, will keep an eye on it.

 

[eli griggs]

This looks to me to be a good marketing opportunity for digi camera makers. It will be interesting to see which company will allow for the option for user loaded software and focus more on delivering quality upgradable sensors/optic platforms to the market at large.

Eli