Hi!
Long time since last post… I’ve been away for a long time.
I decided to share here a little breakdown of a camera mapping job that I did.
I used blender and photoshop.
I hope you like it…
So many good works here on finished projects! This place is very inspiring.
Hi Everton, stunning results. Being able to move the camera around a bit makes the food look 100% more yummy 
! Can you elaborate a little bit on your process? Is it all manual or do you use a 3D reconstruction as a reference?
Thanks, Christian! There’s nothing new or fancy in my process, I just did the regular workflow for camera mapping, I separate the images in layers and reconstructed the overlapped areas in the picture, then I jumped into blender and manually rebuilt the geometry. I used the blam add-on to get the right perspective from photos, now it’s called fSpy https://fspy.io/
The render could be more special and there’s room for improvement adding some compositing stuff here like smoke, dust particles to give more aerial depth…
I’m glad you like it. I also like the magic and the feeling that camera mapping technic brings to a flat boring image…
Thanks, didn’t know about fSpy. The whole process of creating the geometry by hand just sounds very labor intensive. I’ve been experimenting with photogrammetry software like Agisoft MetaShape. You just need some additional photos from slightly different perspectives and you get all camera parameters plus an initial guess for the geometry. I think that might speed up your workflow quite a bit.
Well, yes and no… The photogrammetry is very powerful but sometimes, like in this case where the client have only one image, the “guess” way of modeling is the best and faster solution (the only possible solution since we need more pictures to use photogrammetry), fSpy do a great job calibrating the camera so it’s much more easy to guess the geometry. It’s not meant to be perfect, you just want to bring back the depth so it’s a fast modeling process, since you are with an approximation in mind and not a very detailed modeling result. Other downside of photogrammetry is the dense mesh as result, you usually have to do some retopo or optimize the mesh… If I had the objects (the foods) in hands I woul definitely go with photogrammetry…
