This Fall at ITP I am taking a class “Desert of the Real: Deep Dive into Social VR” taught by Igal Nassima. We learn both the technical and conceptual basics of VR. One of the first assignments is to try out the photogrammetry: scan our own room, and create the 3d model of this space, preferably using Reality Capture.
I have watched a few tutorials where people made photogrammetry look simple and elegant. However, I will need one more week to build the 3d model of my room because the results that I have for now look like a total nightmare.
I initially planned to use my phone and follow these two tutorials: 1, 2 especially the first one for it’s using opensource software: Meshroom and Meshmixer. However, then I learned that Reality Capture has this kind of license that allows paying only for export which was totally fine for me at this stage.
I’ve also installed a few 3d scanning apps on my phone (Cappy, Photogrammetry, turn3d) but all of them seemed more confusing to me rather than helpful.
So, I started taking photos of the part of my room that I wanted to scan and was pretty sure that lighting conditions are fine and I could make just enough pictures to actually create a model.
I took more than 100 photos of a really tiny space to begin with.
After I imported inputs into the Reality Capture, aligned them, and calculated the model, no miracle happened. I’ve got pretty messy results and received them again a few times after when I was trying to narrow down the task and took photos of a few separate objects (a lamp, a clay rose, etc.) in my room — to try to create a smaller model.
First of all, I am going to make MANY MORE photos.
Second, the part of the process after photos have been aligned (components, control points), and before the model calculation is confusing to me now, and I guess that's why the results are so poor. I need to research materials specifically on this part.
Third, I checked out the Insta 360° camera from the Equipment Shop at ITP and will try to use it instead of the smartphone.