2D to 3D Camera Mapping Blender 3D + Adobe After Effects TEST

Another version of the camera/projection-mapped shot of a gas station I did earlier this week.
This time I have more parallax in the shot to show depth. You need to take into account that this is just a single still photograph that has been converted into something that allows camera moves (although limited, I admit).
Rendered/modelled (using basic cubes and planes) in Blender, then composited in After Effects (dirt, grime, lens flare, colour correction, various blurring and extra camera shake to simulate a 'real' camera move).

There are still lots of problems with this shot in terms of getting away with it being inserted into a live action film and not sticking out like a sore thumb:

- The primitive shapes (boxes and planes) I used to construct the 3D scene could be added to in order to create more depth - particularly on the gas station shop which could do with extrusions around the windows and doors in order to add a more 3-dimensional feel.

- The reflection of the gas station on the ground is a bit sad looking. I didn't spend any time on it and the reflection in the previous example is far better (becasue I spent a bit more time on it).

- The Ice Box could be modelled in basic 3D instead of being part of the gas station shop facade.

- I should have rendered a separate depth pass which I would then composite to add true depth of field in After Effects. Instead I just masked out sections of the background in After Effects and faked a depth of field using the camera lens blur effect.

- More atmospheric haze and debris would add depth to the shot and help reduce the feeling the gas station is just a model rather than a real, full-sized construction.

- I drastically reduced the speed of the shot in After Effects to add more 'weight'. I also added little bumpy wobbles to the camera move to simulate a real-world camera person. But I can't help feeling the the shot could do with being slowed down even more.

But this is just an experiement and now I know that camera mapping works and is relatively simple to achieve one the initial set-up is done. Going into The Gimp (or Photoshop) and breaking a still photograph into individual layers based on how they occlude what is behind them, then using the clone tool to fill in the gaps left behind when theose foreground/middleground, etc elements are removed is time consuming (and boring).
Also, setting up the 3D camera in Blender (or any 3D package) and trying to match the angles is a pain in the *%$!

But once that hard stuff is down you star seeing how cool an efect UV camera projection is!