Photogrammetry Beginner to Advanced Starter Guide

Hey,

I have been working on my game World Void for a while now, and it required a lot of Photogrammetry (3D Scanning) work in order to create the assets.

I have written multiple articles on the challenges we’ve faced, and written guides to help people get up and running with Photogrammetry.

If it’s something that interests you, I would highly recommend you check out the following articles:

Everything You Need to Know about Photogrammetry… I hope

practive.png

The title is pretty self-explanatory.

Scanning a Castle Inside Out with No Budget

cameras.JPG

This article is a breakdown of how we scanned a castle and the workarounds we used to reduce the costs.

I also gave a talk about this at GDC and you can watch it on Youtube.

Porting a Real Life Castle into Your Game When You’re Broke

vault.jpg

Once you’ve scanned your object, you want to check out this article that should give you an idea of the different techniques you can use in order to process your scan and create game ready assets.

The Workflows of Creating Game Ready Textures and Assets using Photogrammetry

12.jpg

And here’s a breakdown of creating a gam ready assets, that covers scanning the asset, cleaning the model, extracting the textures, and importing it into UE4.

Alchemy in Photogrammetry

Enjoy :), and if you find these articles useful or you have any questions, make sure to let me know in the comments down below, or you can also let me know on twitter @JosephRAzzam](https://twitter.com/JosephRAzzam). You can also check out my game World Void

2 Likes

Thanks! Night Watcher.

As someone who just started to experiment with this, thank you for those articles!

I do have a quick question: I started with a tree since that seemed to be an easy starting point. The tree itself had some vines wrapped around the trunk that did not come out particularly well when building it in ReMake. So to add more detail could I have taken much closer shots of the vines themselves to help with reconstructing those? or would the software not be able to figure out where they are?

Thanks!

Irkie500: leaves, plants, etc are very difficult to scan. They usually come out as a lumpy mess, even with lots of good, clear photos. It is best to try to remove them from your scene before scanning or when you’re processing the scan, and then add them back in using traditional methods.

Ok that’s what I thought since any amount of wind or blur can really mess with things like that. On a side note have you had experience with scanning branches or the leaf themselves on blackboard or something that you can crop out later? Might be nice to scan a few of those and turn them into masks for later use in Substance Designer for example.

I haven’t done that personally, but I have seen some people do it with good results. One thing to consider would be cross-polarization (as Night Watcher mentioned) to remove the specular shines from leaves. I think you could achieve pretty high quality results that way.

Also, based on what I’ve seen, Reality Capture is very good at capturing thin details like twigs and leaves, compared to other software.

Thank you so very much. This filled in the gaps I had in my workflow for an upcoming project.

Hi, My apologies for the late reply, I have no idea why I did not get a notification about this. but it looks like Wilkes already answered your questions, so thank you Wilkes! I agree with Wilkes about the foliage and leafs, though some people managed to scan them before, but usually it’s not worth the trouble, and yes Reality Capture is better with thinner surfaces, though if you tweak some settings in the point cloud filtering of Agisoft you might get similar result.
and yes for leafs the best way would be to use a blackboard or a blue screen…

I hope it isn’t too late to post here. I’m trying to find a tutorial or guide on extending a photogrammetry scan. What I mean is, I want to take pictures of half of the tree, and model the rest. How would I go about employing textures and whatnot from the scan on the bottom half.

Here’s my theory:
Import scanned model into Zbrush, sculpt tree, UVmap it.
Move into 3dsmax, import original scan, and sculpt. Align the two, bake textures onto sculpt.
Now take the half textured sculpt and just texture the rest of it in Photoshop. Is this a viable method?

The quickest ways is to learn how to let go of the details. I would extract the trunk material from the scan (Diffuse and displacement), then I would just remodel a new try and sculpt back the details use the textures in a software like zbrush. Odds are no one cares if the tree is 100% accurate.

Also in my testing Zbrush was the only software that had a chance of manipulating a scan. also decimate your mesh, there’s always a way to avoid working with such a dense mesh, always.