I took photos of an object with a real camera. I then used computer vision techniques to 3D-reconstruct the object and in the meanwhile estimated the camera pose (locations, pointing direction, etc.). With everything in hand, I imported everything into Blender, and my goal is to reproduce the exact real-world photo with Blender.
There are of course errors in my 3D reconstruction, so by "exact" I mean I shouldn't get "zoomed-in" or "zoomed-out" version of the real image from Blender. Given the fact that the 3D reconstruction of the object is pretty good, I expect to see my 3D model "overlaid" on top of the real object if I align the Blender-output image and my real photo.
During my 3D reconstruction, I set my focal length to 5000, whose actual unit doesn't really matter, as the 3D reconstructed object will just have the same unit.
How should I set my camera parameters, e.g., focal length (how to convert 5000 to Python-Blender's space), sensor size, etc.?
Here is the image whose view (not sure if this is the correct word?) I want to reproduce exactly.
Here is the 3D scene.
Rendering the scene above gives this.
So you see, it's a zoomed version of the real image. This is unsurprising because I don't know how to set the parameters of the camera... Focal length may be easy -- just needs a unit conversion? What about all the other parameters like sensor size, etc.