Depth maps and Facebook 3D Photos from VR180 / stereoscopic 180 footage - Eric Cheng

Depth maps and Facebook 3D Photos from VR180 / stereoscopic 180 footage

Articles

“Automatic” 3D Photo from a VR180 clip. This simple, 8-step process shows you how do to it! 🤣

The process:

extract, rectify/align, find disparity range, estimate depth, rename, upload.

The process, in detail:

1. Download DMAG2, DMAG5 and DMAG9 from Ugo Capeto’s wonderful stereoscopic-photography site.

2. Extract single frames from left and right eyes of the Z CAM K1 Pro clip. You can do this in ffmpeg using this command (sample export of 1 frame, 100 seconds in):

ffmpeg -ss 100 -i input.mov -vframes 1 -q:v 2 output.png

3. Run DMAG9b to rectify/align the stereo pairs (read the documentation–this one is weird).

4. Run Disparity Finder 2 (df2) to find minimum and maximum disparities in the image pair. You can find this executable in the DMAG2 download.

5. Run DMAG5b, enter minimum and maximum disparities, fiddle with window size (I used 16, the default, for this 1024-pixel image), and generate the depth map.

6. Rename the depth map image_depth.png (for original image image.png).

7. Drop the files into Facebook composer, and they will fuse into a 3D Photo.

CELEBRATE!!

I’m sure with many more hours of experimentation, I could get this workflow to produce better results. Alas, I don’t have that time. Most of these conversions work best when artists are involved doing the manual painting of depth, although it’s likely that AI-based methods will improve quickly.

Other methods:

  1. Use StereoPhoto Maker (per Peter Van den Wyngaert’s comment on the Facebook post linked in this article)
  2. Use The3dConverter