Recap is a software program from AutoDesk that manages point clouds. It is provided in Building Design Suite along with Revit and AutoCAD.
AutoDesk have a cloud based service accessed via Recap that will create a 3D mesh file from photographs.
This can be used for objects, buildings or room interiors.
It works by comparing a series of photographs so the way the photographs are taken is paramount to a useful 3D model.
The better the camera the better the result, but any camera can be used.
Fixed lens is best, if using a zoom lens do NOT change zoom setting between photos (move closer or further away instead).
Polarizing filter may be helpful in bright light.
Use a tripod if possible, and remote clicker (because you need to use a low F stop).
Do NOT use a flash.
Use RAW file format if possible, or other loss-less format (e.g. PNG)
Set aperture to be a open as possible (i.e. low F stop), allow shutter speed to vary rather than aperture.
Set ISO to 100, or not much lower. High ISO settings introduce noise (graininess) which is a problem.
Auto-focus can be used but do NOT change other settings between photos.
This process does not work on glossy, transparent and moving objects objects as reflections and changing conditions confuse the software.
Lighting should be diffuse with minimal shadowing. If external photograph early morning or late in the day, preferably when overcast.
Do not adjust lighting between photographs.
Do not allow any changes in objects between photographs (e.g. people, cars, objects moved because they are in the way). Every photograph must include the same objects otherwise the software gets confused.
The background around the object should be messy with lots of things visible. The software relies on things to do calculations, an even monochrome background makes this difficult.
Everything in each photograph must be in focus and not blurry. The software uses everything in the scene, not just the object of interest.
Photograph in a logical sequence. Each photograph must include some part of the previous photograph.
Shoot at 5 degree to 15 degree intervals around the object (depending on amount of detail).
Overlap each photograph by at least a quarter.
It is not a requirement to have the whole object completely within a photograph. It is better to take partial shots from multiple angles.
Move to take photos from different angles, don't take multiple shots from one spot. The software requires different perspectives to do its calculations.
Photograph in sequence around the object, repeat at different heights if necessary. Move in during the sequence to take closeups of areas requiring more detail. Make sure these overlap with previous photo.
If doing an interior room photograph across the room, i.e. stand with your back to walls and photograph across the room - don't stand in the middle and rotate.
Take a LOT of photographs. 70-100 for a small object, 180-200 for a building.
It is not possible to come back later and take further photographs as the lighting will have changed. If the photos taken are inadequate the whole process will have to be repeated.
Check the photos taken and delete any that are severely blurry or have unique things in them like people or cars, avoid deleting more than 2 in a sequence.
If RAW format was used import into an editing program to edit, but make sure EXACTLY the same changes are made to ALL photographs (select all of them before making any changes). Then export all as JPG.
When you get the mesh file you may need to delete the surrounds - they get modelled along with the things you want.