Matterport 3D Camera - Create Levels in Minutes

Interesting, Take a look at ReconstructMe.net also, it's a pretty solid implementation for end user scanning

Take a look at ReconstructMe.net also

Will do! Thanks so much for your post.

Thanks for sharing info on Matterport. I've been starting to get the urge to recreate environments but really overwhelmed at how long of a process it would be. This looks promising!

"ArneOdland" wrote:

Could someone familiar with Unity explain how to import the model to a Unity project?

!'ve done that a few times. For .obj model files jut drag them into your project window.

After it finishes importing (which shouldn't be too long even if it's millions of vertices), you'll need to look at your import options. I assume you'll want to make the vast majority of your imported model a static one, and bake lightmaps, so the important thing here is: if your model didn't already have a secondary uv map on it (which it probably doesn't), you'll need to check the "Generate UV for Lightmapping" option and apply. This is either really fast or really slow depending on whether it is seam-friendly, and sometimes even crashes. If you run into a problem like this, skip generating lightmap uv's and go ahead and drag the prefab into your scene hierarchy, and write an editor script to invoke Unwrapping.Generate-somethingsomething (i forget the name) for each mesh of your model until you identify the mesh that's having a problem. If a mesh has zero vertices for example, that will crash Unity's unwrapping code.

Also, for large model imports you have the option to Generate Colliders for your meshes on import. In my experience, this is very slow through model import options, better off just adding a MeshCollider component to each mesh (manually or via script) after you drag the model into your scene hierarchy.

You'll also need to drag the corresponding texture file(s) into your project window too (not sure if it's possible for obj files to have embedded textures, never seen that). You will likely have to figure out if you can automate the process of going through the model's meshes and setting up their materials with the right shader and texture, or else do it manually.

And finally, depending on how the source model was created, you might have to deal with faces that are sitting on top of each other which looks extremely ugly. For example, models exported from the free version of Sketchup are notorious for this.

In any case, you'll have to tailor the workflow depending on how your source models are created but you should be able to massage your way to something workable. Hope this helps!

Im sorry the quality is terrible. Not useful for ANYTHING but pre-vis and a very rough and i do mean rough outline of things look at the quality of the model in that 3d web demo, not only that but look at the low res of the images . Terrible quality. I mean whats sad is that you can kinda ...its somewhat decent enough to tell what you are seeing at least thats whats sad that you "want it" to be much better but your brain knows its not.

might be good for pre-vis stuff or to give people a "rough" overview of things. but as for something to be used professionally? it looks like crap. Sorrry this thing MIGHT be worth 50 bucks but anything more then that and id say pass.

Some technology things should not come out untill they wait a longer and i say with this they really need to wait until they can provide a) much better models and b) much better image quality. Even with the rifts low res screen i imagine things will feel off with it. Might be worth 50 bucks or so as a novelity item to play around with but anything more then that you'r wasting your money. Because surely something better will come along soon enough.

Nice proof of concept i'll have to say but thats about it.

Sorry. Needs work. Dont mean to be overly negitive but.. the 3d real time demo is really low quality if thats the best it can do it needs work.

" it’s not out of the realms of possibility to see it release at a sub-$500 price point."

Really????? 400 bucks for something like this? No thanks. The quality is very low res.

Please Oculus Rift users please dont buy this becaue what will happen is a ton of people will see this and get one and then starting making demos and levels using it and the problem is it might look fine on a moniter somewhat but if your up close in the rift you will really notice that its low res i bet even with the low res screen. and the fact that all the models are off and not just slightly off much off

This company should have released this as a tech demo. but nothing else. I say wait. please Wait for a year or so until the technology gets better and you can produce something thats worth 400 bucks. This isnt. Sorry.

It would seem that with the proposed specs of the 2nd generation Kinect this tech may become more viable. It does look pretty interesting though the resolution does seem to need a bump before it's consumer ready. I will be following this from now on.

Here's an example of what the above 3d file looks like rendered in Blender.



For a visualisation of a real world 3d space, it's interesting. But for a game, the data is very wasteful. A lot of manual cleaning up would need to be done.

It would be cool for robotics though, an autonomous quad copter doesn't mind if the room it's in looks like it's made from crumpled paper. :)

"kojack" wrote:

For a visualisation of a real world 3d space, it's interesting. But for a game, the data is very wasteful. A lot of manual cleaning up would need to be done.

Yeah the geometry seems unuseable by itself. However as reference this would be pretty good. With the scale and proportions you get it is much easier remodel the space.

Wow, that is quite amazing technology. VR is happening so fast. Sure, the room looks a bit like its made from melting wax, but still would be pretty amazing to scan in your house and walk around in VR. This uses Kinect! Once more advanced hardware comes out, including Kinect 2 likely this year, this product will get instantly better.

We'll be deathmatching in realistic looking historic locations in a few years.

Thanks everyone for all of your realistic comments up to this point. I plan to stay on top of the Matterport developments.

For those of you curious about the cost, I was greeted at their GDC 2013 booth with the comment by David Colleen, "How would you like to buy a $3000 scanner?" Haha. I said, "I'll take a half dozen!" I talked to David for quite a while about the scanner. He seemed to be retired and mentioned something about how he put up the first 3D model on the Internet a while back and got a lot of news attention for it back in the day.

In talking to another Matterport representative, he said, that they would be making about 20 scanners by around July that may be able to be traded in for a later version. They need people to test out the scanners and try new things. By December, they plan to incorporate a new sensor which I don't know much about except that it will likely be from Prime Sense.

Matterport seems to be heavily targeting the real estate community, but I would like to encourage them to fully develop their product to support the ability to "Create Levels in Minutes" like their brochure mentioned at GDC.

They provided a link tonight for the open source software:
Meshlab
http://meshlab.sourceforge.net/

to view and edit the model:
Modest_living_room_and_kitchen
https://drive.google.com/?usp=folder&authuser=0#folders/0B601-8L9cZ8na3hJaG5ydXVSaG8


The main issues appear to be:
1. resolution
2. workflow for cleaning up artifacts and editing scanned 3D models
3. cost

I would imagine that this product will be very viable for everyone within five years, hopefully, in much less time.

Well the real problem is to get a low poly model from the high poly-model they made with their pointcloud. It does not look if they have solved it. Would have been a surprise as much larger companies work on that problem (for example autodesk).
Additionaly the 3D-data from stereoscopy is very noisy. But it is cheap :D