Walkthrough effect with multiple images

  • Hi
    have come across the solution offered by Matterport (https://matterport.com/try-it/) which gives a walkthrough effect with multiple images. Their solution includes purchase of the Pro camera and the software to stitch the images and output the final product for Web. The whole presentation seems to be quite good with a video option. Each room typically has over 14 - 20 exposures giving a typical walkthrough effect.

    Assuming we match the number of exposures and stitch them together, is it possible to get the same effect and output with krpano?

    Any inputs?

    Regards
    GP

  • I don't see why not!

    The zoom effect looks a lot like the RKpano ZOOMBLEND transition effect. You have to make sure that the pano's are correctly referenced to eachother, but then you could easily do this with KRpano. You would have to set a polygon for the clickable area for each pano per pano I imagine to link the pano's, but correct orientation would be automatic.

    So yeah, if you have to time or the automated gear to take a dozen pano's of every single room, this is entirely possible.

  • Good to know this is workable. Actually, I am into 3D rendering (Architectural Visualization) and hence have no issues in rendering any number of panos from the system. So thats not a limitation.

    Ofcourse there is the video part which is an add-on feature. Good to know of the ZOOMBLEND feature. Have not used it so far. Let me try this out.

    Regards
    GP

  • Where can I find out more information / demo on the zoomblend feature as I cannot find it in the documentation.

    Matteport is awesome apart from the costly camera and nasty hosting charges!

    Many thanks

  • I have been looking Matterport for a while now and it seem cool.

    There is one thing I really like a about matterport; you can use the mouse to navigate through the tour and from one hotspot to another.
    Can Krpano use the mouse to navigate like that?

  • I have been looking Matterport for a while now and it seem cool.

    There is one thing I really like a about matterport; you can use the mouse to navigate through the tour and from one hotspot to another.
    Can Krpano use the mouse to navigate like that?


    How do you mean precisely? I only got it working on an iPad, not in any desktop browser, which I wasn't to impressed about.. So I don't know how mouse navigation works, but I imagine that whatever it is, you can't implement the same paradigm in KRpano..

  • Matterport was shiwn to me by my customer. He is very impressed by it. The price quoted by the local service provider is not expensive either (guess to capture the market)
    It would be nice to have such features so that our existing customers are not taken away

  • It looks like a nice complete package consisting of a automated camera platform, photogrammetry software, panorama software cloud system and a viewer platform, but that is about the gist of it. It is a easy to use all-in-one solution.

    For it to be fast and easy, the output seems very limited in features and options. It's not very flexible and therefor only suitable for a number of general applications that require little added functionality and content. It is a completely different tool than KRpano in that regard I think. I could not use this platform for the type of content I need to create. It just isn't flexible enough.

    If some of the features are interesting for KRpano users, I have not seen anything in their demo's that can't be recreated somehow ontop of KRpano, including the photographic model of the environments. In fact, running a couple of pano's through photoscan gave pretty much a similar result. With threejs comming to KRpano soon, that won't be a problem, and right now it is possible to display an iframe with a 3d viewer ontop of KRpano.

    The only real unique thing is the camera which will be a lot faster then out nodal-ninja or gigapan rigs we got going.. On the otherhand, our DSLR pano set is a lot more versatile.

  • What matterport is doing, is not a zoom ... they are making an interpolation in a 3d space....

    You could try to achieve something like this with krpano using the three js plugin,,, but I feel what you get is not worthy the effort you need...

  • What matterport is doing, is not a zoom ... they are making an interpolation in a 3d space.....

    That is not entirely correct. The panoramas are just that, basic flat panoramas. They have no depth information or any sort of mesh. The way they go from point to point semi-fluently is because there are a lot of panoramas per room. if you switch from 1 pano to another, the orientation is know, so the "zoomblend" is used to make it appear that you move through the space, but it has nothing to do with "3d space interpolation", it's just a linked pano tour, nothing else.

    You could try to achieve something like this with krpano using the three js plugin,,, but I feel what you get is not worthy the effort you need...


    If you take a look at the 3d Floorplans, this is photogrammetry model build from all (or a subset) of the panorama images. If you want to have such a floorplan in KRpano, then you'd need to have the ability to project UV-mapped 3d meshes in KRpano. Right now for 3D objects I use another webGL render engine, but the point was that such functionality is not impossible in KRpano.

    But this matterport stuff isn't amazing new tech.. it's just pano + photogrammetry + easy deployment scheme... So the panoramas themselves have nothing to do with 3d models. there just are a lot of them per room, making the transition effect very smooth.

    Whether of not the effort of such a thing is worth it, is for other people to decide. But I do agree that building 14 panoramas for 1 room just for a smooth transition effect is a bit.. excessive.

    Edited once, last by Timescale (May 5, 2016 at 5:51 PM).

  • The only real unique thing is the camera which will be a lot faster then out nodal-ninja


    Not sure about that. As I recall reading from some actual users it could take all day to shoot a house. Not sure how much post processing is involved. The same number of DSLR panos shot with the same attention (fast and systematic) would take me several hours. Automated processing and stitching and you could actually end up with better quality for about the same amount of actual time. It is more like shooting Google Street View.

    Building the 3D model would be my challenge. Photoscan that Timescale mentions looks intriguing.

  • A good setup with a proper camera and a nodal ninja or gigapan will indeed give better more detailed results, as I said, but this camera is nothing like such a setup. The device makes 1 rotation, records the image and perhaps some depth/stereo data and the workflow is seamless.

    If you just want this level of detail, have no technical skills and no desire to add anything more than some waypoints, this system is much faster and easier to handle without training than our basic rigs.

    In fact, I would not be surprised if the capturing and processing of 1 room with multiple, say 10, pano's would be about the same as 1 pano the conventional way.Quality wise, I'd say this sits in between streetview garbage and a proper hi-res pano tour. The multiple pano's per room do make a nice effect.

    Photogrammetry like photoscan provides has been around for ages. I have worked with it quite a lot in my field of work. It has it's... uses!

  • streetview garbage and a proper hi-res pano tour


    That is entirely unfair and perhaps based on a lack of knowledge. Street View is a system. The quality of the results depend on the quality of the supplied imagery and the skill of the provider. While there are plenty of hack Street View photographers there are many great people working with Street View. And there are plenty of hack virtual tour providers, as I'm sure you will agree.

    In my case if a client wants both a Street View and custom website tours I may use the same source material for both. The website tour will get more post processing work and can deliver scads more features but that doesn't mean the Street View tour will be garbage. Example: the Woman's Club virtual tours comparison on my examples page.

  • "That is entirely unfair and perhaps based on a lack of knowledge"

    No it is not. It is based on a deep and founded mistrust of that click-bait rubbish that Google has pushed into the world and the fact that the platform is the most awful piece of crap ever stuffed in the public's face.. It has nothing to do with quality photography and added value by the medium of panoramas. It is pure and utter rubbish and adding your quality footage to it won't make it better.

    Personally I think it is the responsibility of the photographer to explain to the customer how buying into the Google racket is perhaps not the most desirable thing to do and perhaps some of us should grow a spine and sometime tell a costumer that sub-par quality renditions of the work are not acceptable, especially not on a platform that not only abuses the client, but your work as a means to turn all of it into a cheap and expendable b-product.

    Unfair? I do not think so? Lack of knowledge? I imagine I have spend more time thinking about the ethics and problems of this stuff than most... But really, suggesting that street view's "quality" is something worth while and defensible... is laughable.

    Streetview is the bottom end... period...

    Edited 2 times, last by Timescale (May 6, 2016 at 2:54 AM).

  • Majority of our customers are in the Realty sector and they are more than pleased with this solution and the price points. With this setup, any one - with no professional knowledge, can offer solutions to the sector referred. Infact, they are now offering one interior free and charging from the next one onwards (in some cases - even two interiors free). (Just mentioning this, though its purely a marketing strategy). The only point I am trying to make is that there is a focussed approach to push this technology and it is seeing traction in the market.

    I agree with jordi that these are not just images but the process is converting this into a 3D Mesh. As per the team at Matterport, you could even download the mesh after the scan is done. Mail from matterport on the subject

    Quote
    As long as you have a Matterport account, you are able to access and download OBJ files from our portal. You can ownload an industry-standard OBJ file, which will contain both the 3D mesh and textures. This will not include images and panos. So, the obj can be downloaded from any of your models and imported into various 3D programs, including things like Meshlab, Unity, Revit or AutoCAD for manipulation!
    Unquote

    Not sure what three js can do but it would be nice if some one can develop a plugin/code to match these features in krpano. I will be ready to pick it up.

  • "I agree with jordi that these are not just images but the process is
    converting this into a 3D Mesh. As per the team at Matterport, you could
    even download the mesh after the scan is done. Mail from matterport on
    the subject"

    What is there to agree about? A common fundamental lack of understanding the technology behind this?

    What it DOES NOT do, is create panoramic imagery at any chosen point of a room after scanning. The panos you get are the pano's where you took them. These panos are not magic, or holographic star trek technology, just a equirectangular (probably) image.

    The other thing, which is in itself completely unrelated to the panorama, is to take the image and perhaps stereo/depth data to create a textured pointcloud/mesh of the site that was scanned. These models are by definition incomplete because there is no magic "I can see through objects" technology on board because that does not exist. This means that the meshes will have data shadows, AE areas where the scans did not overlap are holes in the mesh. This is inevitable and the reason why the panoramas seen are NOT the result of a 3D rendering.

    I mean, If you had a perfect 3D model textured and all, why not simply put it in the Unity engine? You could walk around fast and jump just like a 3d game! It would probably be a hell of a lot more efficient on the server side.

    the 3D model provides you with a neat floorplan that you can spin around and roughly see some details on, nothing more. If you want a 3d model of room, you can do that now if you want with your normal camera and a piece of software.

    The only "new" thing about this method is that the pano generation and the 3d object (photogrammetry) creation has been streamlined into 1 process so that people with limited experience in these matters can still get results. But there is nothing magical or revolutionary going on here.

    Three.js is a javascript/HTML5/webGL 3D engine. Klaus already has a experimental demo somewhere where he demonstrated 3d animated objects as hotspots. Another engine that is quite interesting is Babylon.js. I have actually used that to display 3d meshes of rooms that I scanned and processed with photogrammetry software.

  • Hi Scott
    checkout your examples page. Nice work and a good feature comparison between Google and Custom tour. Question: When you say you use the same source material for both: How do you get access to Stret view source material?

  • Hi gputhige,

    You can download Street View panos using a Chrome plugin (Pano Fetch). But that isn't what I do.

    I shoot my pano images and batch process the HDR. Then I batch stitch them in PTGui. If any problems show up I can fix them manually but that is rare. I then upload the finished sphericals to Google as "pre-stitched". In this way I bypass Google's less capable tools and maintain quality.

    For the custom tour I may tweak the images such as manually retouching the source images or adjusting the HDR blending and of course adding the nadir. I can easily spend as much time perfecting one pano as it takes to process the entire tour automatically.

  • This is a great forum for debate as we can all see that the likes of Matterport are indroducing great conecepts which bring nice results which we all want to replicate and keep the costs and "profits in house" so the the more we can work together to find an alternative through our own endeavours the better. At the end of the day Matterport is trying to target the end user "The real estate agent" which it cutting out part of the chain so we have to be more forward thinking to provide something even more interactive.

Participate now!

Don’t have an account yet? Register yourself now and be a part of our community!