When you say "develop" a raw file, do you mean convert it to a viewable colour space, given a known screen? Like HDR or some other type of choosing the exposure/gain?
Re: 3D, are you aiming to infer a 3D scene from a single photo? There's a few approaches for that, but they tend to show off the examples that work well, and conveniently avoid examples that work poorly. But depending on the content of your photos, you might find an algorithm that works out of the box.
For data representation, we used a sparse voxel representation of a Truncated Signed Distance Function to represent surfaces for this project: https://www.microsoft.com/en-us/research/project/kinectfusio...
However, I can't recall interacting with the sparseness; that level of abstraction may have already been resolved in the code by the time I got on that projects. TSDFs are cool as hell and way more interesting than representing ambiguous surfaces with mesh triangles.
Also, I've been there with the turbulence a couple times. Sometimes months happen in a day and vice versa ;)
Re: 3D, are you aiming to infer a 3D scene from a single photo? There's a few approaches for that, but they tend to show off the examples that work well, and conveniently avoid examples that work poorly. But depending on the content of your photos, you might find an algorithm that works out of the box.
For data representation, we used a sparse voxel representation of a Truncated Signed Distance Function to represent surfaces for this project: https://www.microsoft.com/en-us/research/project/kinectfusio... However, I can't recall interacting with the sparseness; that level of abstraction may have already been resolved in the code by the time I got on that projects. TSDFs are cool as hell and way more interesting than representing ambiguous surfaces with mesh triangles.
Also, I've been there with the turbulence a couple times. Sometimes months happen in a day and vice versa ;)