Sorry your browser is not supported!

You are using an outdated browser that does not support modern web technologies, in order to use this site please update to a new browser.

Browsers supported include Chrome, FireFox, Safari, Opera, Internet Explorer 10+ or Microsoft Edge.

Geek Culture / Wow - digital photos into 3d models? Cool!

Author
Message
NeX the Fairly Fast Ferret
19
Years of Service
User Offline
Joined: 10th Apr 2005
Location: The Fifth Plane of Oblivion
Posted: 1st Aug 2006 21:55
Found this by accident:
http://news.bbc.co.uk/2/hi/technology/5235724.stm
Could have many uses.


Since the other one was scaring you guys so much...
BatVink
Moderator
21
Years of Service
User Offline
Joined: 4th Apr 2003
Location: Gods own County, UK
Posted: 1st Aug 2006 22:25 Edited at: 1st Aug 2006 22:27
Microsoft reinvent the wheel again!

This stuff has been around for years, I did one of one of missus' cottage ornaments back in 2002...

http://devink.pwp.blueyonder.co.uk/cottage/cottage1.htm

Once it's loaded (~300KB), you can rotate it in all directions.





Attachments

Login to view attachments
Jeku
Moderator
21
Years of Service
User Offline
Joined: 4th Jul 2003
Location: Vancouver, British Columbia, Canada
Posted: 1st Aug 2006 23:58
That is not the same thing

This new one will allow you to take a single image and transform it into a 3D model. The closest thing I've seen is from the University from here that had software analyze a walkaround video in a room and create a 3D model from it.


"I understand creative people. After all, I worked with towel designers." - Ray Kassar, former head of Atari
BatVink
Moderator
21
Years of Service
User Offline
Joined: 4th Apr 2003
Location: Gods own County, UK
Posted: 2nd Aug 2006 00:06
The write-up says you have to take multiple photos, which it then analyses for similarities and constructs it from the differentials in the picture. Which is how that one up there ^^^ works



Dazzag
22
Years of Service
User Offline
Joined: 26th Aug 2002
Location: Cyprus
Posted: 2nd Aug 2006 00:07
Quote: "This new one will allow you to take a single image and transform it into a 3D model"
Wrong tech. You are thinking of something that was posted a week or so back. Not bad effect with a train and a street scene. This is totally different. I posted it on the cool list thread earlier today. There is 2 videos on the MS site. Basically it is a way of browsing by images. The end idea is to be able to, say, view you images of a place and then seamlesslyish splice it together with other peoples images from around the world. So you get a more complete image. If, for example, you took a picture by big ben then the program would splice in images from around your picture to show a much bigger picture. Nicely if someone took a close up of something then you can zoom in from your picture (or one of the others) onto the close up. One of the videos sort of suggested this could be like browsing on the web (eg. the close up takes you to a website dedicated to the item in question). The 3D part is sort of how it shows what it thinks is the complete scene. Sounds pretty cool. The videos are pretty good.

Cheers

I am 99% probably lying in bed right now... so don't blame me for crappy typing
Jeku
Moderator
21
Years of Service
User Offline
Joined: 4th Jul 2003
Location: Vancouver, British Columbia, Canada
Posted: 2nd Aug 2006 00:11
Oops, my mistake! Sorry


"I understand creative people. After all, I worked with towel designers." - Ray Kassar, former head of Atari
Dazzag
22
Years of Service
User Offline
Joined: 26th Aug 2002
Location: Cyprus
Posted: 2nd Aug 2006 00:23
No worries. But check out the videos. Most impressive if it can be pretty much seamless. Bit like Google images with a massively better interface. They way they market it is you set up the indexes and it can work out where your image goes in a line up of other images. If it works well then we could have a Google images type tool with almost perfect virtual tours of anywhere on earth (most famous the better obviously for more images). Plus, if you think about it, then you would never need a powerful camera. Especially for really famous places. I mean why bother with more than megapixel or two when someone else would have taken a close up picture of each part? Cool.

Cheers

I am 99% probably lying in bed right now... so don't blame me for crappy typing
Matt Rock
19
Years of Service
User Offline
Joined: 5th Mar 2005
Location: Binghamton NY USA
Posted: 2nd Aug 2006 05:24
They tried this at my university, but they were terrible engineers lol. It ended in utter failure and it's awesome to see that someone is finally succeeding with it. Wow, can you imagine the implications this will have on the game industry? 3D modelers might be out of a job


"In an interstellar burst, I'm back to save the universe"
Dazzag
22
Years of Service
User Offline
Joined: 26th Aug 2002
Location: Cyprus
Posted: 2nd Aug 2006 10:21 Edited at: 2nd Aug 2006 10:24
Quote: "someone is finally succeeding with it"
Well, hopefully. We will have to see how good it is at automatically identifying nearby images. Plus it will be interesting to see what happens when millions of people upload their pictures. I mean it would only be awesome if it were on a large scale (ie. lots of people), but this could also have major problems. Imagine wanting to do a tour of the area around Big Ben. Unfortuantly part of the clock (yeah, yeah, it's the bell, whatever...) matches a certain other picture indexed under "Big Ben" that has nothing to do with famous landmarks.... More interestingly if it can match shapes etc to work out where images meet then might be more embarressing if an image looks like something else. Naked statues for instance. I'm guessing some sort of human QA factor here. ie. you can't just post and link any old thing. You have to wait for it to be authorised.

Quote: "3D modelers might be out of a job "
I doubt it. Watch the videos and that BBC article is pretty much way off on the idea of creating 3D models with it. They just use a 3D representation of all the images. Some bits look quite 3D, but overall I don't think so. Saying that software has been around for ages to produce models of buildings from a few shots. If you have hundreds or even thousands to work with then would be easy.

Cheers

I am 99% probably lying in bed right now... so don't blame me for crappy typing
dark coder
22
Years of Service
User Offline
Joined: 6th Oct 2002
Location: Japan
Posted: 2nd Aug 2006 10:40
i find it very hard to believe the program could just search through millions of images and find one that fits, it would surely require some human categorization or editing, and in that video all the examples seemed to be from a single viewpoint, so if you had photos from various angles it just wouldent work well, but i would have to see it to believe it.

Hallowed are the ori.
Dazzag
22
Years of Service
User Offline
Joined: 26th Aug 2002
Location: Cyprus
Posted: 2nd Aug 2006 11:03
Yeah, there has to be an element of human interaction. Sort of skimmed the video but they show a bit of that. I'm guessing that the computer will make a guess and then the user can move the image around a bit. Perhaps even some clever AI to do "snap to" type logic like some applications do with multiple windows. Just a lot more complicated. eg. you move the image towards it's proper location and the computer then realises that line A in image 1 is line D in image 2 so moves them together automatically. But this will probably be all behind the scenes. ie. you submit your big ben pictures then a few days later they are all seamlessly added to the virtual tour. Or something like that.

Imagine if they synced it with Google Earth. Now that would be amazing. Scroll over to the UK, then zoom into London, look around a bit, then zoom right in on Big ben and get a 3Dish virtual tour of the area. Right up close. Tops.

Cheers

I am 99% probably lying in bed right now... so don't blame me for crappy typing

Login to post a reply

Server time is: 2024-11-17 10:28:09
Your offset time is: 2024-11-17 10:28:09