Sort by time | Sort by thread (beta)

Link to this comment:

Share to:
see all

All Comments (248)

Sign In or Sign Up now to post a comment!
  • COOL!!!!!!!

  • please, some information of how you develop this software? do not give false hopes

  • Wouuuu!!!

    

  • How do I get this? :) Could I e-mail you possibly? PLEASE! :(

  • Great work! This system is amazing and has many potential. I hope that this project will grow and will make its way into popular software repositories :)

    Have you made this or do you plan on making this work open source?

  • Way too cool. One PhD for you, served. :D

  • open source for 3d printing!!!

  • Comment removed

  • Serious question: Why is the scanner not picking up the desktop or the hand?

  • @Graceeeeem It works out the structure from motion, as the desk isn't moving, it gets ignored by the programme. The hands don't conform to rigid body motion, so the programme can't track them effectively, hence they are also ignored.

  • how do i get this software

  • 12 people are gay.

  • Yay! Now we can rip things from Real life: The game rather than Halo and COD!

  • There are some glitches but this is kool.

  • What happend with this?? still in development?? i the webpage doesent apear any download link or nothing, and its been more than a year since y posted here, and was posted in the blendernation webpage.

    Pitty if the proyect didnt go trought. I was really a great thing.

  • GENIOUS!

  • @viktorgv thx for your info! :)

  • ImageCV based?

  • When can we get this software?

  • do u have software as a trial or beta ?

  • @karandex vi3dim

  • WOW.. This is brilliant! WOW!!!

  • This is really great.

  • Fuc*ing Cool

  • I have a invetion like this to tack it to the next level

  • @sakkurae

    This is not a product demo. It is an academic demo. He will be releasing code (look at his website) but you are perfectly able to download the paper yourself, read it, and write the code yourself. Not sure why this is a "scam."

  • The video makes it look easy, thats the point of the software.. so how easy is it in code form? Its not. I want to download it and install it, with a GUI to play with it - I essentially want a release candidate/version build or beta software.

  • @caturn88 you want the moon on a stick from the sounds of it. Try doing some work yourself for once instead of letting everyone else carry your talentless ass

  • @IPUNCHEDALION oh a yank, thanks for that. Your the type of person who doesn't get things. That's right. I shouldn't be bothering but I have the time today. I don't need to counter your argument, because its crap, but I will say that a demo video must be of a product that is o be released in working order for people to use.

    This very product is to make things easier for people, so where is it, try try.

    Who ever doesn't get this is surely stupid.

  • You assume that its value lies in the product it produces instead of the furthered knowledge in the area. And demos are coming Real Soon Now according to the site.

  • @qwasdninja Thanks for your constructive reply, what you say is no doubt true but from a video like that the purported software is going to be lust after by many.

    I just want to use it, releasing demo software is normally a natural part of something like this, but then this is a Uni project.

    I checked the site and its still 2009 since it was last updated.

  • i read about this in computer magazine and thought, thats a useful program........ make it availeble fast!11 people can even PAy For It!1111212

  • so where can i download this. if i cant download this, this is a scam.

  • hola me gustaria descargarme este programa , me puede decir donde descargar, si me pueden contestar el privado .gracias

  • One more thing. It appears that there is no averaging of assumed polys from frame views. I watched the wrong assumption triangles from the steeple to the arch disappear and reappear even though new information from the video would provide averaging leverage to keep them gone.

    The major poly groups also did not try to "eat bigger" the smaller, but in plane polys to reduce overall poly count (thusly allowing the model to become less accurate in some darker spots the longer it was modeled)

  • Since the object is a "Rigid Body" I have another suggestion.

    Include a "Lighting variance option" too.

    If a desk lamp's spot light is waved over the rigid modeled object and the program is aware (by entering another modeling stage) that lighting changes are happening, then some useful triangle refinement information can be in play.

    For example, reflectivity info and bump maps could be added for improved texture. It would add in some false triangles with shadows, but that can be averaged out.

  • Since stray vertex triangles are a problem in the modeling, how about a "Squeeze down" factor?

    For example, if the person holding the church model had moved a few sheets of different colored paper (white, black, yellow, blue, green, red) behind the model then the stray triangles could've been "squeezed to the nearest surface because it would be obvious that they are not part of the scene. At the moment the model is too triangle-stingy and resists removing early wrong assumption triangles.

  • imagine that program for making titles.

  • This is really neat technology. Unfortunately it would be very impractical for doing real modeling, since the polygons are constructed very poorly and in a very haphazard manner. But as someone else said, this could be great for future robots that can learn and map out new environments in 3D.

  • @CapnDingo The key here is online. It might be impractical for use in a video game, but it is much faster than typical laser scanning and other more accurate techniques. I think that there are plenty of applications (like robots) for this sort of technology.

  • This would be so nice to try out, I can't wait for those months to pass when it'll be made available for testing.

  • Or you could read the paper and implement it yourself...

  • SO your telling me I could make it for myself? How come they have a mailing list for when (in a few months) they release software for Linux and Windows??!

  • @caturn88:

    I'm a bit puzzled as to what having a mailing list and a software release has to do with having (or not) access to the information and being able to implement it yourself.

  • (I'm not saying it would be trivial to implement, especially with the tweaking required to get it running in real time, but you could definitely get something up and running by yourself)

  • @ostrorawr its really simple, I want to use the program for myself (probably like a lot of people) but I'm not a developer so need it compiled and with a GUI... I just want a download link.

  • @caturn88 You can download the paper, read it, and make it for yourself... yes. The mailing list is for people who don't want to bother with writing it themselves, not because you couldn't do it. This is an academic work, so the technique is freely available in the paper on his site.

  • @ostrorawr:

    Why don't you?

  • @Spectralancer:

    I will next semester in our 3d geometry reconstruction class.

  • You're going to develop a whole program?

  • GTA in your own city !!!

  • where is the sourcecode?

    freesource for the win!!

  • @juaxix You don't need the source code. The details are in the paper.

  • uhm... sample source always helps :)

  • Hmm, great idea! Impressive :D

  • Forget games...this thing's true potential lies in robot vision. If computers could generate a 3D environment out of a simple camera input, one could, for example, write a program to drive a car in a city. That's the end of traffic jams and accidents.

    The possibilities are endless...

  • There IS a program to drive a car in a city! I saw it once, don't remember where!

  • If you just want a good 3D model, then this probably isn't for you, but imagine it's gaming potential? You could show it you mobile, and the use the tracking feature to use it like a Wii remote! Good work!

  • @jeffwinterbourne HAHAHAHAAH stupid idea

  • So what if it isn't a clean model? It has great potential and not every application of 3D models needs those models to be clean.

    Bravo!

  • If this is the best demo they can come up with it has serious flaws. The joints where the tower meets the building has loads of remnant polygons across the hinge. Considering the simplicity of this shape would love to see how it copes with more complex shapes

  • @JonRashid its an academic project FFS!

  • Check Out Photobuilder from Autodesk. The fact its academic is irrelevant - some of the leading developments in this area are from the academic area wtf

  • Quite impressive! Great idea.

  • Our congratulations!

  • Shame it would not be much use in gaming environment - or much else as the need for correction and tidying up would take longer or at least as long to model it correctly in the 1st place by a reasonally competent 3D artist... Good work tho...

  • Not much use in creating high poly models is what you probably mean. I think it has great gaming application potential. Imagine if a kid wanted one of his action figures in his game.

  • Obviously you dont have any idea how game models can be. Even low poly models. this demo is good but not good enoght for any production even games

  • hard work indeed

  • i hope you get rich....

  • Linux users are going to love this. Good luck on the Windows Version. This is a freaking showstopper in my eyes!

  • Rename it to: World of Warcraft: Behind the scenes.

    Very nice tho :O i wonder what would happen if i tried to do that with my stuff

  • agreed on the freeware! looks great. i would love to use this for soo many things from game development and web design to even custom viualizations for music preformances.

  • Extremely impressive! How does it know not to track your hand?

  • you have to relese it freeware!!! PLEASE, if so only just for linux!!! this is REALLY AMAZING!!!

  • Freeware and Linux? In that case just open-source it.

  • that was what i was trying to say, but didnt get the right idea in the comment, thanks, that was, make it open source for linux users.

  • wouldnt 2 cameras enhance the quality of 3 dimension (like the case is with human eye)?

  • Absolutely AMAZING

  • Really Impressive !

  • Comment removed

  • I have the same question as bakacdaz has.

    I am highly interested in using this software(?) for architectural models.

  • Cool..!!

    But...

    Does the Resolution of webcam effect the model.

    I mean if i use expensive webcam is my work quality will better than cheap one.. or it the same

  • This is awesome! Great work! :-)

  • Very neat. well done.

  • there are still little bugs around the roof/tower area. perhaps from a bug in the delaunay-surface reduction? but good work so far! :) I look forward to seeing what this will result in.

  • amazing!

    Open source This Please!

  • That would be awesome! open source please perhaps as libperforma! I wonder if it could even be integrated into programs such as pidgin for improved low bandwith video conferencing?

    The point it is if it isn't open source how can progress be made from it without developers spending loads of money? Maybe you could sell your ideas to nokia and have them integrated into the QT toolkit?

  • yes, where's the code already?

  • Molto interessante nell'ambito dell'architettura. Potrebbe funzionare anche su un oggetto immobile e la camera invece che si sposta?

    Grazie

    Arch. Emilie Rollandin

  • This is freaking awesome dude! Well done!!

  • of course some work got to be done but it's alredy fantastic

  • awesome, got no any words. flawless

  • Excellent work with great potential! Thanks for posting the videos.

  • How come the hand was not rendered at first? It's good it didn't happen, but how?

  • you notice he didn't place his hand in front of the model... it probably would have picked it up if he had...

  • @1Madis89 look at the hand, every time the object turns around it changes it's position, so is not added. this is what i would say...

  • Where can i download this cool thing :D i can't fine the download link ind theys homesite

    sorry for bad english right now xD

    PLEASE SEND A PRIVAT MESSSAGE WHIT LINK XD

  • wonderful.

  • It's amazing! Can't wait to see how this kind of technology will be used in game and Google Earth etc.

  • Pure awesomeness.. All the best for your PHD :-)

  • Where is I can download this soft?

  • This is FANTASTIC!

    If the model was shot against a green screen or something similar and rotated using automated rigging could you avoid having to augment the video? Ie automate the whole process.

  • This could really change technology as we see it. You should really capitalize on this man, imagine the offers you could get from Microsoft, or Sony.

  • I think this is a fantastic idea! I've done modeling before, and this would greatly cut down the amount of time spent to build object models. And it's also an amazing use of a webcam. Well done!

    As for the naysayers, the question shouldn't be "What's the point?". It should be "How can this be useful?". This type of 2D to 3D live modeling is not the end product; it demonstrates what's possible, and may one day become a vital part of something that many can not do without. (like mobile phones?)

  • simply amazing!

  • Comment removed

  • i definitely agree about the possibilities for the future, and don't want to bring down your work, so keep at it!! The tech will certainly come about in its own time...just consider pushing the idea sooner than later and you will be very fruitful with the tech, especially if the intention is marketability. Perhaps coordinate with those you may view as end users (studios) so they will support the tech possibilites

  • 3D scanning has been around for years and modelers continue to do things by hand, as the models that come from 3D scanners are usually too feature-bloated and too high polycount to be really useful expecially for animations.

    This would be a really good way to do a "quick and dirty" 3D model that the artist can use to develop a cleaner model.

    It's fun how a church model was used... were you citing the classic Paul Debevec's work? :P

    PS: wish my PhD would have been so fun! Keep up the good work!

  • I would like to see a higher end 3d model...at this point, it's a novel idea but realistically, the polycount is silly considering the geometry, and any professional model/texture artist can create that in right about 30 minutes max. So, besides saying "neat"...whats the point?

  • How about creating a photorealistic MMO?

    Matrix here we come!

  • Maybe if would be even more amazing if he tried to model his hand, but I guess is too advanced for a "beta"

  • This is amazing... next step, reduce the un-needed polygons. Dunno if this would be optionable live but a markup of the 3d object afterwards could do the trick.

  • good work Qi! ^^

  • it's really awesome :D

  • I'm really looking forward to piracy in 10 years.

  • Sure. The thing is if they make a commercial lite version of this tool even none 3d modelers may be able to create basic 3D models for stuff like home videos.

  • Amazing. As it does such quick tracking in 3D space, could this system not be used to quickly make an 3D input device?

  • good stuff qi - it's all about the cents.

  • Just imagine the accuracy with a Natal style depth sensing camera. Apparently, one of the guys who used to work at 3DV systems is now at Microsoft and works in 3D environment reconstruction.

  • I would like to know when this would be commercially available as I know many, including myself who would like to start using the application.

  • A demo should be released in a few months depending on my PhD research - the problem is what makes interesting research doesn't necessarily make the product more user friendly and vice versa.

  • No doubt.

  • The most impressive thing to me is how robust it is at interactive speeds. Exceptional work.

  • Nice work. Interesting how the algorythm seems to ignore your hand. I suspect you manually selected the region in the video prior to passing it to the recognition algorythm.

  • The hand doesn't rotate on the y axis like the model does, so most of the recognised landmarks on it would probably be discarded. Also, if you look well at the beginning, part of the hand is recognised (on top of the roof) but then the algorithm removes [most of] the unneded polygons.

  • This is very useful and can have variety of applications. correct me if i'm wrong but augmented reality only scans flat pictogram which must already exist in database. This feature scans full 3D object, catches transformation in all axes. Try to do that with flat pictogram.

  • great stuff want to see it on a big scale camera moving and objects static this would make a great tool for the gaming industry imagine whole cities with realistic textures and you playing

  • This is the coolest thing I've ever seen.

  • Amazing. Finally an inexpensive rapid modelling tool.

  • Way to miss the point..

  • QiPan - You are genius :)

    This program would do wonders for archeology students in the world - imagine it used to catalogue objects in museums - you could rotate museum object from Cairo on your computer in Cambridge...

    Imagine archeologists scanning dug out pieces of broken objects and then putting them together in computer.

    This program will make archeologists and museums salivate. Go to Cambridge museum and show them what can you do just walking around a statue with your laptop in hands

  • Congratulations very beautiful technical exploit!

    1 / question: is the placement of points on visible edges or on surfaces?

    2nd question: where let us can test this software ?

    Congratulations also to Delaunay and Voronoï ;-)

  • oh yes? show me that

  • wow toy companies will have boost in sales for sure.

  • I just watched this video on Engadget. It looks cool, but it's not obvious what the applications are - apart from looking cool...

  • I'm assuming this isn't yet marketed... any ideas as to how and when it will be available??

  • where do we download the program? *Serious face*

    ...

    hehe i want the software so i can create models for a game =) will be fun! 8D

  • dont get too excited, look at how many errors in geometry were made, and that model was basically 2 rectangles. Imagine trying to scan in a more complex model, you'd spend more time clearing up errors and fixing the needless verts.

    Its impressive, but 3d-scanning has been around for years, and its just not viable for that line of work. It would be ideal for creating displays.

  • the point is 3D scanning need a particular 3D scanner and a fixed turntable. Base on the video, Qi's concepts is a software which can generate 3D object using webcam only. It is alone a brilliant idea.

  • Seriously impressive. Would you be able to get this running in AS3 using the papervision3d library I wonder? That would be an incredibly useful tool.

  • Is there a camera calibration step, or can it do this without correcting for lens distortion, etc?

    How much impact does the camera resolution have on the quality of the results?

  • The camera needs to be calibrated for this system, otherwise there would be an ambiguity during reconstruction. Higher res would mean higher quality, but might not be able to run in real-time, as higher res = more features = more computation.

  • Could you run it in non-real time?

  • RAD

  • Very interesting. How does the algorithm distinguish between the object you want to scan, and other objects such as your hand, the desk, etc? Does it look at other frames and take the difference as the object you want to scan?

  • The hand, like the background, is discarded from the reconstruction as it does not conform to the rigid body motion of the object. The only assumptions made are that the object is a rigid body and at least part of it is within the white circle you see at the start. No background subtraction is being done, which means that the camera does not have to remain completely still (e.g. can be worn on an AR headset)

  • Absolutely epic!! I can already think of so many things to do with it!!

  • you can bet an updated version of this technology will be used in the sims 4 in 2014 to make your character look exactly like you haha..

  • can't wait for its public release!

  • Wow, this is really cool ! keep up the good work and write us an easy to use software !

  • Brilliant! Definitely useful for augmented reality as well.

  • excellent work, you should work a bit more in the hand, it still looks 3d. Maybe the textures are not accurate. The desk doesn't look realistic at all.

  • Amazing Mr Qi Pan,how old are you?

  • Thanks - I'm 24

  • Very Impressive work. On what kind of computer did you run those tests? Was it normal "citizen class" hardware, or any members of the "supercomputing league" ?

  • Thanks. The system you see in the video was running on an Intel E6700 2.6GHz dual core cpu with 2GB of RAM, so just a standard desktop PC.

  • Brilliant. I'm sure there's a huge science side to this, but you've just made 3D easy for the film industry.

  • Not quite yet - this technique only works for textured (detailed) objects for now, but we are trying to improve the algorithm soon to be less dependent on texture. This is only a proof of concept after all!

  • right, but people can take actually sculptures, or puppets and manipulate them in a computer. most of the time a physical replica is created before hand. you've just found an easy in between step.

    now the question I have is, once the scan has finished can you break apart the 3d model into layers for separate manipulation? Also what kind of file does it save as? is it compatible with other 3d programs?

  • Generally speaking its quite hard to fake a live demo at an international conference (and then win best demo prize for it), especially when also reconstructing things from people's pockets such as packets of chewing gum, cans of beer and mobile phones. However, I am honoured that you think my results are good enough to be fake ;-)

  • Awesome work, very impressive.

  • Well this is a start, for sure. Once the algorithm gets better, this will be kick ass.

    Amazing work!

  • This new process is amazing!!! As a computer science teacher in high school this type of thing will build students interest and fire them up to continue in the field.

  • Wow x 2! Can we get a copy of this software? Assist with beta testing?

  • Wow this is amazing, Im just wondering how you cancel your hand out of the model building process as your hand is part of the scene that is changing. And If someone with a different colour hand used it would it still work as well. I'm glad no one saw this video (or mentioned it) before my honours presentation last week, lol as I did something similar but no where near as good. Congrats with this amazing system.

  • The system doesn't actually need to take into account the colour of the hand to get rid of it. The hand, like the background, is discarded from the reconstruction as it does not conform to the rigid body motion of the object. The only assumptions made are that the object is a rigid body and at least part of it is within the white circle you see at the start. Hope that explains it a bit! Qi