Modeling images from a 360 degree camera

Development Asset Creation
,
1

Hi there (first post go easy on me).

I’m looking to model static content (like a room in a house) from a series of 360 degree images.

Will RealityCapture be able to model accurately from them? Or is there some way to turn the 360 degree images into a form that RealityCapture will support?

Thanks,

1 Like
2

Spherical images will be supported vry soon. Hopefully in next release.

However, it is not supported in current release.

3

hi there,

can you give us an indication when the new version with the 360 degree support comes out ?

btw … awesome software :wink:

thanks,

4

Hi Florian
There is no hard ETA, but I can say it’s almost ready for deployment…

5

This feature would be very helpful.
I’m currently generating a model from 360 video footage captured by a GoPro Odyssey camera with this workflow:

  1. Convert stereo video footage to mono
  2. Convert equirectangular footage to cube map
  3. Export each face of the cube map as a separate image sequence
  4. Generate model with RC using cube map faces.

It’d be great if RC could directly process the equirectangular images.

6

Great that they are adding the feature for picking up scenes via video!

I have a 7x Heroes rig that could come in handy for some shoots in weird scenarios, but since I’m not extracting my images from video I just set each camera to take a pic every few seconds. Only done it a few times in testing, but good for a quick capture of tree lines strung from some strong fish line, or on a 15 dollar dolly/skateboard to grab things like church pews if constrained.

I’m not doing 360 video anymore, so that set-up is more of a re-purpose, I’d be curious as to the results of a scene from 360 video, what with invariable seems you’ll have to some degree. (Though brilliant for cube-map usage)

, how is the GoPro Odyssey? How’s the stitching? Does it require a lot of finesse like the original 360Heroes rigs?

7

Is this close? it was almost ready in November. That’s a quarter of a year now, a full billing cycle for those paying every 3 months.

8

Any news on a future update that allow camera alignement with 360 images ? we plan to buy a istar, it would be a great functionnality for photogrammetry workflow…

9

hi there,

you are implementing that 360 image feature right ?
we will make 360 with the gear 360 from samsung - its just hobby but when it work we will get a istar for the company.

Can you please give us indication when the feature will be available.

thanks!

10

Hello everybody,
please excuse the delay. This release depends on various in-house factors and we cannot tell you the exact date of the release.

11

Hi, any news on 360 camera support? It would be great if we could use these.

12

hello CR Team,

May I ask for a short update about the 360 view feature?

We really wait for that feature - we want to use it to capture fast and very special areas on our hospital construction projects.

Thanks for Info!

BR,

13

Just adding my name to the request list.

Please support equirectangular images as input.

Thanks!

14

So, I’m converting the source 360 images into 6 flat images and it is not really working.

Should I instruct Reality Capture in any specific way, so it understands that those images are related? I’m getting a lot of small pieces and linking in via Control Points would be a nightmare.

Please help,

Thanks! :slight_smile:

15

It should work if your stitched pictures was not too distorded (in a way non solvable by the lens model equation).
Can you provide a sample ?
You can try to do a draft alignment and see if this is better.
You can also try to do cube faces that overlap (more than 90° field of view).

16

set your distortion model to division.

it will probably give you the best chance of matching them.

17

for every three months is fair I think …

 

___________________

Institutes in Lucknow

18

Excuse inexpert question - how can photogrammetry use pics from a single 360o scan position? I thought it all worked by parallax from multiple positions, unlike laser-point-cloud which records direction-plus-distance?
I do understand that part of RC’s speed is because it converts the data into direction-plus-distance but it must convert that from initial parallax data?

19

It can’t. You need multiple 360° images taken from several positions.

20

Pleased to find my understanding wasn’t wrong.

So it’s a bit misleadsing when in latest Jul/Aug 2017 AECMagazine ‘Leica BLK360 preview’ it says
“once the button is pressed 3mins later you have a ful 3D scan of the room, in a combined laser, photogrammetry and heat-mapped model that is ready to be used instently in Adesk Recap Pro from a partnered iPad”

Jonathan_Tanant wrote:

You need multiple 360° images taken from several positions

Does that mean just as many positions as if single conventional photos were taken, with same overlap guidelines etc?