Hi. I am trying to model a car but i dont have it’s blueprint.
What i humbly did was i took two pictures from two angles. Then i use BlenderTsai.py to locate the cameras back in blender. Next, i divide the screen and set the reference photos for both cameras. Then, i just draw the the car shape on left screen. After finish drawing on left side, the car will be coplanar. What i did is i use the reference photo on the right screen to recover the depth of the photo on the left side. I did this by extruding all the vertice and put them all at the camera’s center. By scaling the extruded edge, i can control the vertice so that it wil project exactly at z-axis(view axis) of the left side camera (i like to call it zz-projection).
The problem i am facing is i cant make the vertice overlap on exactly the same spot. I think it is due to my camera’s barrel distortion. I think this is a nice way to model without blueprint but with the problem it cant produce accurate model.
Can anyone try it out and improve it?
Or do i need a laser scanner?
Blender is not a CAD but a visual tool. Even if it were there would be a certain amount of distortion depending on your lens, distance from the subject and so on. Blueprints are orthographic, reality is not.
It sounds like you’re making too much rocket science of this. Just relax, have fun and make up what you’re unsure of. Merge to centre those points that are neither at point A nor point B.
Thx for the quick reply. Maybe i am too perfectionist. I cant bear the feeling to make it lay at the same spot on both reference photo. The problem is myself, I hate this.
hmm i dont know much but i would just say take a piece of paper and a pencil and draw what you want your car to look like left/right side, Front and back and top all on paper each different view on a piece of paper then scan it on the computer and use that as a background image to base on what you want to make your car model look like or in your case pictures from left/right side,front&back, top
I think there’s still a program called ‘photomodeller lite’ that can make models out of multiple pictures. basically you load in the various views, and then you select points to ‘tag’, and you do it with all the views. once you have entered the point data, it cranks out a dxf model that you can import into blender. assuming you entered the correct value for the camera lens, your model will be a good replica of whatever was in the photos you worked from. I tried it out with a toy truck, and it worked great.