ImageStitch was nice functionality added in v13, but is there anything available for constructing meshes (with color textures) from a bag of 2D images (taken at different angles/positions)?
You can reconstruct surfaces (with as little as 10-20ish photos) easily on some mobile apps like widar, and many python implementations exist for this problem that are efficient, most prominently NGP from Nvidia.
I would like to know, what has been done on this topic for wolfram language?
To get started, here are a few simple test sets containing 10s of jpegs showing an object at different angles with uniform backgrounds: https://github.com/AlansCodeLog/photogrammetry-test-sets
Here's the code to get the first one, which is 33 images of a rock:
URLDownload["https://gitlab.com/photogrammetry-test-sets/dice-turntable-strong-lights-heavily-textured-w-watercolors/-/archive/master/dice-turntable-strong-lights-heavily-textured-w-watercolors-master.zip","~/Downloads/"] files=ExtractArchive[%, $TemporaryDirectory]; images=Import /@ %
Here are some links to the research on Photogrammetry and a few relevant tools:
COLMAP is a general-purpose Structure-from-Motion (SfM) and Multi-View Stereo (MVS) pipeline with a graphical and command-line interface: https://colmap.github.io/tutorial.html
Original paper/repo on the first method capable of photo-realistically reconstructing a non-rigidly deforming scene using photos/videos captured casually from mobile phones using neural radiance fields (NeRF): https://nerfies.github.io/
Nice video about photogrammetry with examples for archeology: https://www.youtube.com/watch?v=1KJNCOJAULw
Repos with useable code: