Why haven't computer-game animators tried this yet?

Have computer animators ever tried to build 3D models from actual photography? Here's what I mean. Let's say I'm interested in creating a World War 2 fighting simulation game. Instead of creating a tank model from scratch, I could simply find a real-life tank and take a picture of each dominant side panel and "slap it on" a wire frame model that I've made on my computer. Circular features of the tank--like its firing barrel--would be a lot trickier, but flat planes should be pretty simple to paste to.

Of course, it would be best to photograph this real-life tank on a gray day, to prevent shadows, sun reflection, etc. Either way, the model would require a lot of post-production work, but I feel the final result would be far more realistic than how we build graphic models from scratch.

So imagine a first-person shooter game where everything you see is made of crystal clear photographs of real-life objects (or at least planes). Is this game design approach even feasible?

Not only is it feasible, it's exactly how 3D game modelers create many of their resources. They simply take photos or concept art, map them to flat rectangles and position them in Maya or Max so that the plane of the quad they're mapped to is perpendicular to the line that the photo was shot along. Then it's simply a matter of setting up cameras to face these quads, switching between them and using these background images as guides for positioning vertices. It's cheating in a way, a bit like tracing 2D art with tracing paper, but it works well and in practice it's used a lot.

yes it can be done (blender) but no it is not being done as a photograph is a bitmap and games rely on vector graphics - vectors are created mathematically so they can be manipulated accuratley and quickly

Because photographs slapped onto 3D models look dodgy?

Also, you are simplifying 3D art way too much - unless they are boxes, they have more than 6 sides therefore more than 6 photographs would be required.

Here's what happens in some 3D games:
1. Create 3D model
2. UV map determines where the 2D art goes on the 3D model
3. Apply the colours and textures to the 3D model

The 2D art wraps around the model, it needs to curve unnaturally in order to look natural once applied on the model. Have you ever extracted 2D skins/textures from a 3D model? They can look pretty weird sometimes. (I occasionally work with skins from The Sims 2)

I doubt all videogame art is made from scratch, at least for 'realistic' games - Many artists use photographs as guides and some might even take colours and/or textures from them.

There are a number of super-realistic videogames out these days, so it's not really needed anyway.

But yes, my point is, that it doesn't work very well so it's not done.

It's called image mapping, and it's been done since the beginning of computer animation.
I remember doing it in 1992.

It looks crappy if that's all you do. It's obvious.