incendiary69
01-31-2009, 03:46 AM
Hello, I am starting to dabble in 3D modeling, and since I have no real experience (and arguably not much artistic skill at this point), I thought I would start by relying on reference photos. What I would like to do is model a simple shape and then use a 3D Paint program to "projection paint" from a photograph in order to texture the model.
So I have a picture of a cereal box taken from an everyday perspective. I have no information about the exact position, orientation, or focal length of the camera that took the shot.
I'm comfortable approximating it as a simple cube, and I have the relative dimensions of the box, so the initial modeling is not the issue. My problem now is how to align the model with the original photo, so I can paint the texture on properly. I have the original photo textured to a plane, and I want to position the model in front of the plane so that the model occupies exactly the same position as the cereal box did in the original photo. This way I can clone the texture from the plane onto the model (or "stencil" it -- I have access to several different 3D painting programs -- BodyPaint 3D, Mudbox...)
This is where I am having a lot of trouble. My understanding is that from a single photo, if one didn't know the dimensions of the object, one would never be able to say for sure what its position in 3D space is. But I have the dimensions, and I know, for example, that the top of the cereal box forms a plane, and that this plane is perpendicular to the side of the box, and so on.
I can obviously move, rotate, and scale the cube. I can also move and rotate the camera, change its field of view, etc., which has a different effect. I just need some help conceptualizing this so that I can get it right. I'll rotate it along one axis until it best fits, only to need to rotate along another axis, which ruins the first fit. And don't get me started with movement along the Z direction, which changes the overall "shape" of the perspective distortion.
Is there something I can do to help myself out here, some assumptions/constraints about the knowledge of the photo...? Is there some way to align the coordinate system of the painting program to the plane of the cereal box as it appears in the photo.... or something like that...?
I hope I've explained what the issue is. Any insight would be greatly appreciated...
Thanks,
Steve
So I have a picture of a cereal box taken from an everyday perspective. I have no information about the exact position, orientation, or focal length of the camera that took the shot.
I'm comfortable approximating it as a simple cube, and I have the relative dimensions of the box, so the initial modeling is not the issue. My problem now is how to align the model with the original photo, so I can paint the texture on properly. I have the original photo textured to a plane, and I want to position the model in front of the plane so that the model occupies exactly the same position as the cereal box did in the original photo. This way I can clone the texture from the plane onto the model (or "stencil" it -- I have access to several different 3D painting programs -- BodyPaint 3D, Mudbox...)
This is where I am having a lot of trouble. My understanding is that from a single photo, if one didn't know the dimensions of the object, one would never be able to say for sure what its position in 3D space is. But I have the dimensions, and I know, for example, that the top of the cereal box forms a plane, and that this plane is perpendicular to the side of the box, and so on.
I can obviously move, rotate, and scale the cube. I can also move and rotate the camera, change its field of view, etc., which has a different effect. I just need some help conceptualizing this so that I can get it right. I'll rotate it along one axis until it best fits, only to need to rotate along another axis, which ruins the first fit. And don't get me started with movement along the Z direction, which changes the overall "shape" of the perspective distortion.
Is there something I can do to help myself out here, some assumptions/constraints about the knowledge of the photo...? Is there some way to align the coordinate system of the painting program to the plane of the cereal box as it appears in the photo.... or something like that...?
I hope I've explained what the issue is. Any insight would be greatly appreciated...
Thanks,
Steve