Thorskan – Better Reality 3D Scan | Hacker News Reader

Thorskan – Better Reality 3D Scan

geeks3d.com

32 pointsivoflipse22 comments

Quite prominent grafiti in the first demo says: "F*ck Cracow's police" but in polish and without asterisk.

We are not very PC in Poland.

Seems like 2014 might be the year of photogrammetry, between this and the Microsoft thing posted a few days ago.

There's a lot of VERY impressive photogrammetry software out there - see http://en.wikipedia.org/wiki/Photogrammetry - but it hasn't really made it into the mainstream yet.

Incidentally, one of the big uses for this tech will be the heritage / museums industry. They're very big on photogrammetry for capture and display.

I can't imagine this technology is directly adaptable to video games. These videos seem to be pre-rendered which isn't going to help in video games. You won't be able to kick any of that trash around or break a window. It might be helpful for architecture purposes depending on how it is implemented. I guess I'm just not super-impressed without seeing how it is done.

There is already a game in development using this technology - "Get Even" by Polish studio The Farm 51:

http://www.youtube.com/watch?v=L48SdXmd0_k

http://www.thefarm51.com/index.php?module=news&id=50&menu=1&...

I guess I just don't see what the advantage is from creating the environment manually. Sure, if you know of a warehouse that already has great graffiti and the exact layout your game needs then go for it. But what if you need a stairwell or a hallway? Are you going to build it? Or have to find one and then modify it so it matches the photo-realistic nature of the process? I'm sure there are level designers out (and I know that there are 3D artists out there) that could easily recreate this. Usually in-game they are given limits that prevent them from doing something like this though.

This would appear to be the same environment as the demos from the article. Note the 'Node' graffiti on the wall, and the general shape of the room.

I'd really like to see a before/after comparison, to see what the original environment looked like, as well as an actual description of what they've done.

I'm guessing it's very hard to capture a scene in daylight and then modify that model with different lighting schemes.

How do you know what's the color white vs a very bright light source?

Maybe the user can indicate light sources somehow?

What I found most fascinating was the degree to which detail increases when the entropy of human constructs increases without intervention. Eventually this aspect will reverse and it will all become homogeneous. What actually is this thing that first increases and then decreases when a highly ordered system progresses into disorder?

Closenes to optimum of your perception that happens to be between "boring" and "so much chaos you can't figure what you are looking at".

like between Universe in the beginning - hot meson-gluon soup and at the end - vast over-expanded vacuum with few super-super-super-large black holes

TBH, this kind of technology was what I was hoping Photosynth was going to turn into.

I work on photosynth, so allow me to respond (albeit with a bias).

Photosynth was designed to work with the most common use case - pictures taken with cell phones at different, undetermined angles, with 10-20 seconds in between each shot.

Contrast that with the camera system used for THORSKAN - http://i.imgur.com/VE1qXwS.png

That's 72 professional DSLRs set up to take pictures of the object simultaneously. I'm not discounting THORSKAN's accomplishments, these are some extremely beautiful demos. But they can't be created by your everyday user.

Eventually things like photosynth will get on this level, but the camera technology in phones will have to change. Luckily, this is already happening. Intel's CES announcements focused on stereo camera systems in laptops and portables, and MS has an investment in it with the kinect already. Having even two cameras shooting at the same time dramatically improves results.

Hi thanks for the response. I can appreciate how having consistent equipment at known locations and angles taking images of a scene can help with scene reconstruction. But I'm not sure I follow on the need for very close temporal alignment. Is it because of the consistent lighting conditions? With all things being equal, would it matter if each imager used with THORSKAN took a picture 20 or 30 minutes apart?

When you say "very close temporal alignment" do you meant the stereo cameras he was talking about? If so, (and I'm sure since he works on Photosynth he is infinitely more qualified to comment but this is the internet so I'm going to anyways) then the advantage of a stereo camera is that you can more accurately judge depth as long as you know where the two lenses are in relation to each other. Think of it like like those old-timey 3D images. [1] It appears as though Intel's version is going to be a single camera with a depth sensor like the Kinect.[2]

[1]http://en.wikipedia.org/wiki/Stereoscopy [2]http://www.display-central.com/free-news/display-daily/3d-ca...

Not exactly, he made several mentions of the important of multiple shots happening within a very tight time frame (or at least that's what I was reading into the comment).

Temporal alignment matters because, well, things move. This isn't as much of a problem for vast concrete structures that are indoors like in the preview they have here, but in the outdoors things are shifting all the time. It's hard to reconstruct something in 3d if it's changed shape by the time you've gone to take a picture again.

Not going to lie, saw the videos and my first thought was "Get out of here, stalker!"

I first wanted to link to their homepage, but its just a sign up page for updates.

Their Vimeo has some more nice looking videos like this one: http://vimeo.com/60564990

Too bad they aren't exactly spilling beans on how their tech works, what it takes to run or any details for that matter. I guess we'll have to wait, unless someone else has more info.

I can't wait to have it on Google street view.

Apple acquired C3 Technologies which does 3D Scanning from aircraft.[1] It is pretty awesome technology. Google has their own version I believe. I haven't heard of Google Street View vehicles using laser scanners for depth readings though.

[1]http://9to5mac.com/2011/10/29/apple-acquired-mind-blowing-3d...

AFAIK, post Thrun, all Google Street View vehicles have a laser scanner(s).

http://commons.wikimedia.org/wiki/File:GoogleStreetViewCar_S...

Makes sense. Especially since they worked on that Radiohead music video that was "shot" with laser scanners.

Reply on news.ycombinator.com