3D models built with Kinect style depth camera

[dacke93]

Ahh that's how they created a realistic Matrix!

[dextanity23]

@ThereRLG2006

Yes, but what is the bigger part of that shift is combining this with ubiquitous eye/head (in the form of glasses at first, contacts eventually) mounted stereoscopic(obviously) Augmented Reality solutions. People are not even remotely realizing how dramatically that technology (and all the infrastructure that goes with it: distributed cloud computing, much faster internet, etc being necessary) is going to reshape our society, and it may only take a decade or two to get there.

[partikla86]

what kind of software allows you to connect kinect to pc and walk by and make meshes out of it???

[tomasusan]

@partikla86 This kind of software.

[MrLegendman3]

It looks like shit. But it is the best looking piece of shit I have ever seen.

[kokocipher]

I'm really impressed

[RGBDvision]

Look for the technical paper titled "RGB-D Mapping: Using Depth Cameras for Dense 3D Modeling of Indoor Environments"

[gonzo110391]

@RGBDvision

Sir, how do you translate the data obtained to 3D CAD model? Strictly speaking, what sofware do you use?

Thank you. Really appreciate your work.

[EwanColdicott]

What did you use for localisation of the depth camera? I've tried to do this before with a Kinect + 6-DOF IMU but couldn't get a result this good as couldn't accurately determine the position and attitude of the camera. Each point cloud on its own looked great, but it became a mess one I started translating each point cloud into a global coordinate system.

[cristohv]

just put this on robots and BAM!!...smarter robots will be able to see the world like us :D

[06cobaltsc]

I cant wait for the day we can create our own 3D maps for FPS games and play them on pc or console with our friends. 10 years off which is nothing. I should run to the patent office.

[DJBonc]

did the lab release any source code or binaries that were involved in making this video?

[zppz]

it looks like you are just rendering colored points, am I right?

[bradkey98765]

@ThereRLG2006 Don't get me wrong, but so long as the ability to move and see slight different angles of something, is thought as lesser than having a much better "3d" image, It's not going to take over just using 2 cameras. However the kinect is showing great signs for making software to track body positions, which can then be transferred in real-time to a robot or virtual "character". Maybe in the future mmorpg's will use this technology and online worlds will become allot more interesting.

[ThereRLG2006]

@bradkey98765 Well, obviously, the Kinect itself, as it is now, it's just too low-rez. It needs to be able to scan 5 to 12 Megapixels, and right now, I think it does 640x480, which comes out to 0.3 MP. But this will change HOW people think about Photography. Right now, you take 1 picture, 1 view, 1 fixed perspective, and that's all that picture shows. You can take 100's of shots, but that only sorta gives you an idea of what location looked like.

With this Tech, you can actually BE THERE.

[bradkey98765]

@ThereRLG2006 Well the kinect still limits, e.g. A kinect in front of a person will not capture the back half of them, and then leave a "shadow" of uncaptured space behind objects. "With this Tech, you can actually BE THERE" , I suppose with the use of virtual reality and pre-scanning of the area done on a kinect like device. But I was mostly against the proposition of using the technology on phones.

[ThereRLG2006]

@bradkey98765 I'm not talking about Kinect, as it is now. I'm talking about what will happen in the future, using this technology to make something BETTER. You gotta think ahead here, see where things are going.

For example, let's say you open a picture on your phone, TV, or computer. Just imagine grabbing the edge of that picture and changing the camera to where ever you like. Have you seen Microsoft's Photosynth? Picture that technology on steroids!

[bradkey98765]

@ThereRLG2006 "For example, let's say you open a picture on your phone, TV, or computer. Just imagine grabbing the edge of that picture and changing the camera to where ever you like", Well a picture from any 3d sensor will always have the limitations of not knowing what is not in view e.g. the back of your head. You will have very limited abilities when it comes to rotating an image, unless you take around 4 pictures simultaneously around the object.

[bradkey98765]

@ThereRLG2006 If you say that later technology will be better and will be able to capture all angles of a an object from a single view, the technology will be completly different to the kinect and that will be the "AMAZING leap forward". As far as the tech goes it limits the ammount of rotation by quite a substantial amount, the maximum distance is very limited, The technolgy isn't flexable enough for "photography" in a point & click there's a full 3d model way.

[ThereRLG2006]

@bradkey98765 Is it your intention to comment me to Death? Moving a robot without some kind of force feedback is a dumb idea. Go look up Haptic Controllers, that's the future of robot control. And as far as justifying your stance against Holographic scanning based on the limits of modern technology, that's as dumb as saying, back in 1998, that we'll never see video on the internet because of 56k modems limits. Things change with time. You type too much, and think too little.

[bradkey98765]

@ThereRLG2006 Hmm well really I think the biggest outcome is a cheap motion tracker for animators. A single kinect like device no matter how high quality will never be able to capture a full 3d model of anything in motion, and it would be impossible to capture complex objects like a tree without literally going fully around the tree at all heights, and that's if you assume no wind and that some plants emit infrared which may cause interference.

"photography" to me isn't indoor 3d modelling.

[bradkey98765]

@ThereRLG2006 Or at least if you mean for better calls, there's simply no reason to have them, because a 3d video call would be more effective by the use of 2 cameras because you simply don't need any more than one view in 3d, and it doesn't even seem that 2d video calls have taken off. But If you didn't mean for video calls... and more of a camera. Then for still images it will only be useful with head tracking technology or a new display system like holograms that do not exist today.

[bradkey98765]

@ThereRLG2006 I personally like to think of the things that are already possible with the kinect and what will become possible given software development. Like my example controlling robots. Using the kinect as the method to control a virtual character in a virtual world, with full 3d and head tracking, which have both been done and time will only improve the software. We have a very long time to wait before the tech shrinks to be a handheld device, and holograms is even further away.

[ThereRLG2006]

@bradkey98765 I know people love the movie Minority Report, but I think body tracking is probably the LEAST useful part of the Kinect. My Wii is collecting dust, and I know it's not alone. Motion interfaces are novel, but they aren't really that useful. They are exhausting to use, and lack of the precision of a mouse and keyboard. Also do you really want people to see what you're doing well playing a video game? And most of us don't have cool, graceful moves... think Star-Wars Kid...

[bradkey98765]

@ThereRLG2006 I was thinking more for example controlling a robotic arm with precision that is otherwise an incredible task to achieve on a mouse/keyboard, using a device that also has many other applications. I am certain that tracking human motion accurately has many possibilities, games are one, but i do lean towards mouse/keyboard. But then there's ai in robotics, without a depth sensor artificial navigation is somewhat impossible.

[bradkey98765]

@ThereRLG2006 Well I don't know, I just feel that there must be many useful applications for body tracking, not just the cliché software controls. But I'm sure someone who has an interest in AI in robotics will find the use of body tacking and depth sensors now commercially affordable to be a huge leap. The question of which aspect will change the world the most is a matter of opinion, and the best agreement is the applications are many, most of which are very exciting :)

[bradkey98765]

@ThereRLG2006 Being as we don't really see in 3d, we simply see two 2d images, in terms of photography and video it will always be far more visual to simply use two high resolution images. Why would you put this technology into phones (An infer-red projector, an inferred-camera and a RGB camera), when you could instead use two cameras which from a users eye would look like a better 3d image. "Holodeck" sure but we don't have any cost effective hollodeck projectors?

[MonokelJohn]

Really really cool! I am looking forward to be able to easily scan real 3d objects and saving them on my computer.

[soulofsounds]

Is there any software like this commercially available? I'd love to use a Kinect to make 3D models of objects or interiors/exteriors. It would save a massive amount of time compared to taking photographs and stitching them together.

Also, how detailed can you make the textures?

[aburahamamu]

これでバイオハザードやってみたい。。



[sniff699]

excellent work!!!!

[NIPPONNNNN]

すげえええええええええ

[leoleonardo]

@ThereRLG2006 I agree. I personally love to think of the Police/Military applications of a system like this. Police can 3D map buildings. even Firefighters of the such, Having instant access to full 3D walkthrough style imagery with out huge capture equipment and large computers is amazing and could save lives. Not long till we see "Ghost in the Shell" Esq Things. 

[rbneville]

@leoleonardo am i the only one who remembers seeing something like this device on CSI...they used it to scan the crime scene and do a digital reconstruction

[AuraQualic]

10 more steps once we have this in standard video camera we will see movies in REAL 3D not in face 3D.

[badmephisto]

Well done! Wish there was more detail. What prevents you from constraining to one room maybe, but getting a huge amount of data to fill in all the holes and make it look very pretty?

[lievenvv]

beautiful! absolutely brilliant man... i just got one today, can't wait to get my first point clouds in. check out what ill be doing with it! youtube.com/watch?v=oxjGRL7ZiL­E

I'd love to learn how you registered the 3D point data btw!

[DoyenOfDestruction]

This is amazing, I wish this had more than 700 views. Videos like this delve deep into the possibilities of sensor technology and its future potential. Bravo!

[frederickfarrell]

that is brilliant! I bet movie producers would love it, take it into an amazing building and map it so you don't have to spend hours digitally recreating areas for CGI shots. Obviously better quality would be needed but in time that should come I'm sure. the promise for augmented reality is amazing, imagine a contact lens or glasses that give you a HUD that combines what the kinect sees or has seen. you could play with created 3d objects like they were there or in any environment you wanted!

[xhable]

What's the filesize of that result?

[aiklarung]

@xhable Probably not big considering it's a huge bunch of squares. Surely less than 1GB