Fotowoosh, a new service from Maryland-based startup Freewebs, will turn any image (preferably an outdoor image) into a 3D model. They went live on Friday. Examples of what the service can do are above (along with the original 2D images. A video is here which shows more examples.
The 3D image is constructed in Virtual Reality Modeling Language (VRML) format, meaning you currently need a VRML reader to see it (future browsers will likely build this functionality in). In a week or so, the company say, users will be able to upload a picture and have a 3D animated image returned to them in a Flash widget that can be embedded on any website.
When you upload an image to Fotowoosh, their software tears it apart and distinguishes the sky, ground and vertical elements within the photo, then cuts and folds it into a 3D model:
Our system automatically constructs simple “pop-up” 3D models, like those one would find in a children’s book, out of a single outdoor image. The system labels each region of an outdoor image as ground, vertical, or sky. Line segments fitted to the ground-vertical boundary in the image and an estime of the horizon’s position provide the necessary information to determine where to “cut” and “fold” in the image. The model is then popped up, and the image is texture mapped onto the model.
This is the creation of Derek Hoiem, a PhD candidate in Robotics at Carnegie Mellon University, who’s now working with the company. Additional information on the intellectual property behind Fotowoosh is here and here (these links auto-download a pdf and a powerpoint document).
Microsoft is working on something related to this in their Live Labs group called Photosynth (more information here). The product will construct a 3D model based on lots of photos of the same thing or general area from different angles.
Freewebs raised $11 million in venture capital in August 2006 from Columbia Capital and Novak Biddle. The company’s main product is a website building tool that draws 18 million or so visitors per month. Shervin Pishevar, the company’s president, say that Fotowoosh will be a standalone service, and they’ll also integrate it with offerings from partners as well as the Freewebs service itself.
This has great applications for showing houses online. It’s hard to get a true spatial awareness of a room or lot without 3D - this could do the trick!
Wow now that is cool! And a neat name too
Am I the only one getting motion sickness from looking at these pictures? From the sounds of things, if this gets widely adopted, I might have to start taking some Gravol before browsing the internet.
I wish them luck, tools like this are often fun to play with but struggle to find good applications.
Another candidate for “they did this before, and possibly better” is Mok3, Inc. They have wonderful tools and won the DemoGod for 2003, got written up in Ny Times, USA Today, etc. They were lauded as the most impressive demo by many folks.
http://www.usatoday.com/tech/c.....side_x.htm
They even did some great work for the city of Boston to do 3D projections of the big dig. But i guess it never quite got traction as a general tool — looks like they have reconstituted into a 3D travel site.
http://travel.supertour.com
VRML lives! Done a bit VRML markup a couple of years ago.
This would be a nice feature in Adobe After Effects. You can do it already in After Effects, but you have to break the elements out into various layers and position them in 3d Space. This looks easier, but who knows what they do on their end.
I want this for my ‘curvy’ friends photos on Facebook
It’s certainly hugely impressive: I can understand why this might be useful for visualizing older photographs, but… surely it would be more practical to just, y’know, wander around your target location with a video camera, authentically capture the scene, and then compress and upload a Flash video of that?
If I could have that as a desktop background across both my monitors I think I could brag a little more.
Rock on, CMU!
thats cool…. a picture essentially worth a thousand books.
Nice trick for industries wanting to show a new perspective on a 2D image, but we all know the difference. Take your train photo and let my fly the camera under to see the back side of the wheels…. say there is brake trouble for example and you are doing a news story animation. A cute trick, but not 3D.
So just my 2cents - this is really impressive and super cool. Shervin kudos to you guys for taking a stab at something this complicated. obviously anyone versed in these matters realizes this is quite novel - they don’t use photogrammetric tools since this is done only from one point of view. i understand (read the papers Dip linked to at the CMU website) they use something dubbed geometrical contexting where they basically assume that every photo has a 3D model consisting of sky, flat ground surface and vertical planes. they then analayze the phto and define the various regions and then basically define the bending points and morph the image onto the “assumed” 3d model - Shervin was that close to correct?
Also i wonder if the process is indeed fully automated…i would assume that depending on the input image even the smartest algorithms could render silly results….if the Beta SW is client only and with no user intervention (read: they dont do any slash tech touchups in the server side) i would be even more impressed!!
at any rate one wonders what would be the mass appeal of this technology -
* Consumer level (Fliker)….well it seems that the main forte of the technology is well defined geometrical shapes. notice that the fine structure is lost (see trees and so on). Beyond that i assume that people will be distorted in this type of rendering (see the demo with the street and person standing on the left sidewalk)…and since most common photos people take and store have people in them this could be a big hindrance for wide spread sustained use.
* Marketing - definitely. super cool effect that is probabaly very well suited to a world where 3D cameras are not yet abound!!
* Real estate….again. big IF IMHO. as some people mentioned not only is fine detail lost here but there is very limited immersion effect and some distortion. Although this is indeed cool i am not sure what the added value will be over current solutions in this field like video tours, panoramas, flash and so on. there misght be a backlash against the lower preceived quality that these techniques yield (something like “no, no, dont worry the fireplace isn’t a trapzoid…oh and that black patch in the right corner is just a rendering articaft”. But i guess time will tell. obviously these are very talented guys and this is just preBeta.
*gaming - interesting indeed. i guess only the creators can say what kind system requirements rae needed (MEM and MIPS wise) and how far this is from OTF level.
Shervin - I wonder if you would care to comment? Especially on the human distortion issue.
Mike
very nice!!!
this is very very interesting
不错
sweet!!!
This thing is amazing. Wow. I wonder how they do this.
that is so cool
This is really cool
WOW!!!
Thank you for your site
I made with photoshop backgrounds for myspace or youtube and whatever
my backgrounds:http://tinyurl.com/6rbxmr
have a great day and thank you again!