How I 3D scanned the Treasury at Petra without leaving home

Published on

January 9, 2019

After reading a random article about the Petra Treasury, I felt a strong urge to 3D scan it (it’s both my hobby and profession). I decided to do an experiment.

How to 3D scan an ancient structure on the other side of the world without leaving home:

Step 1: Run google image search for “Petra Treasury“, set size filter to 4K and higher.

Step 2: Install Firefox add-on called “Google Image Downloader

Step 3: Download 200 random Petra images from google.

Step 4: Delete images that are not of the Treasury, or are bad, blurry, etc.

Step 5: Import remaining images (about 100) into RealityCapture (Photogrammetry software)

Step 6: Align images

This is the point where I was genuinely surprised. I wasn’t really sure if this would work or not, and this was much better than I ever hoped for.

Step 7: Generate Mesh and clean up a bit (delete edges and remove noise)

Now I’m actually grinning like an idiot. Again, this turned out much better than I imagined.

Step 8: Generate Texture

Viola. Not the greatest scan in the world, but the fact that it worked at all is very amusing.

Here is the 3D model if anyone wants to check it out.


I did a similar trick with Mt. Rushmore. I extracted 400 frames from a drone video I found on Youtube.

Here is the 3D model.

I use ClipGrab to download high resolution video from YouTube. It’s free but be careful, their installer is very sneaky with installing 3rd party junk on your computer. Make sure you opt out of everything it offers.

EDIT: I have switched to using this online tool for downloading videos, it’s the only one I found that can download resolutions higher than HD:

I then use Free Video to JPG Converter to extract every 3rd frame. This left me with about 400 pictures for each scan.

The results are pretty good. Here is the model:

The Peggy Cove lighthouse is the subject of one of my favorite photos:

Image may contain: cloud, sky, outdoor and nature

Again, about 300 pics extracted from a drone video:

And the resulting model:

Update Jan 18:

This is probably my best “found footage” scan so far: