The Canon EOS R5 Can Now Shoot 3D Images With Standard Lenses

1) The dual fisheye is not a "regular lens".
2) IDK why it would need dual pixel? You already have a stereo lens attached, so it already records two stereo pair images. Should be quite easy to generate a 3D image from that, even without the Canon software.
 
Upvote 0
They better bring this to the R5II...

I can't imagine they'd try to charge for it, at least not the basic functionality. Maybe the cloud rendering or something like reality capture does through. Here's hoping.
 
Upvote 0
That is true and the whole point is that Canon developed a software that gives you a similar effect with a regular lens that is not a dual fisheye.
No, gotta call this one out. The article and the one they linked don't do a good enough job distinguishing or discussing this.

The dual fisheye gives you a stereo image. One you can use with VR/AR to represent the entire scene to an audience as 3D. The distance between each lens aperture is similar to the average distance between pupils.

Dual pixel RAW cannot capture depth for a full scene (it's not a light field camera). And can only give you a depth map within a working range that is going to be different for each lens used, subject, distance to subject, settings used, etc. And it requires more processing. That processing let's you build a model with a mesh and texture of the usable area that is then viewed.

With the dual fish eye you can just display the 2 images or video through a headset (or get creative and generate a cross eye).

As for the overlap capturing process necessary for photogrammetry the dual fisheye lens is useful because you get 2 images of exact known projection. Makes processing more accurate in many cases.
But it's wide angle fisheye too... So an entire other can of worms to deal with for photogrammetry.

As it stands you can capture dual pixel raw and use those images in photogrammetry workflows already. But that's without the special knowledge that they're dual pixel, and without the correlated depth map. It's treated as very highly overlapped images instead - still useful!
 
  • Like
Reactions: 3 users
Upvote 0
There was some buzz around Dual Pixel when the 5D IV came out, but ultimately these are not new techniques. As @andrewmckay mentions this is basic photogrammetry, which has been heavily used e.g. in aerial photography and the like.

I found this nice animated image from the 5D Mk IV 2016 here: https://www.magiclantern.fm/forum/index.php?topic=17695.msg171832#msg171832

And just a bit further down theres a computed depth map: https://www.magiclantern.fm/forum/index.php?topic=17695.msg172097#msg172097

Ultimately how useful it is in a consumer context depends on how easy the software is to use and how the depth information is actually used.

When the iPhone X came out with portrait mode in 2017, this was basically what Apple did. They just hid all the implementation details and named it portrait mode.

While Canon just put the raw Dual Pixel feature in their cameras but never did too much with it.
 
Last edited:
  • Like
Reactions: 1 user
1) The dual fisheye is not a "regular lens".
2) IDK why it would need dual pixel? You already have a stereo lens attached, so it already records two stereo pair images. Should be quite easy to generate a 3D image from that, even without the Canon software.
You missed the point entirely. You need AF, and that is how DPAF works. The point is that you do NOT need a "dual fisheye stereo lens" attached. No need to purchase or rent that dual lens at all. The 3D images being generated with a normal lens. You are NOT generating 3D images with a normal lens without software.
 
Last edited:
Upvote 0
This could be a new direction for Canon cameras. After all, all RF cameras are equipped with Dual Pixel Autofocus (DPAF) functionality. It's truly amazing to see this feature now being used to create 3D images. I've heard that back in the 5D Mark II era (Maybe) , third-party developers had already experimented with the DPAF feature to create some interesting effects—I think it was something like "single-frame HDR," right? But perhaps the R1 might not be able to use it, because some of its pixels are split vertically instead of horizontally... Haha.
 
Upvote 0
The first paragraph needs to be rewritten, it says that the dual fisheye is a regular lens.
I hope the DPRAW will be enabled in R5II with firmware upgrade. It helped me occasionally with previous cameras, and I had it always enabled unless I need the maximum fps.
 
Upvote 0
Article needs to be rewritten...
"using regular lenses like the Canon RF 5.2mm f/2.8 L Dual Fisheye 3D VR."

"When viewed on a screen, the image could be rotated, revealing its three-dimensional nature. This capability rivals the 180-degree 3D modeling of Canon’s costly Dual Fisheye lens, but at a fraction of the cost"
Currently, to achieve any sort of 3D output from an EOS R camera, you need a lens such as the Canon RF 5.2mm f/2.8 L Dual Fisheye 3D VR. and that generally runs between $1500-$2000."

If the software only uses DPRAW and not the dual fisheye lens then it would be the only useful application of DPRAW in post that I am aware of.
 
  • Like
Reactions: 1 user
Upvote 0