COMP5421 Project 2: Single View Modeling
Author: Chang Zhangyu (zchang) & Li Yuliang (cs_lyxab)
The Pentagon
The Pentagon is the headquarters of the United States Department of Defense. Here we construct this building by using 5 source pictures, one of which is from
Google Earth and the other 4 are bird-view pictures from internet.
Source picture:
Source Picture 1
Source Picture 2
Source Picture 3: From Google Earth
Source Picture 4
Source Picture 5
Texture
One texture, which is the ceiling of the Pentagon, is obviously non-quadrilateral object. But the texture file itself is still a rectangle. We removed the
unnecessary part by making them transparent.
The VRML Model
This is a still image of a new view of the reconstructed scene.
Corridor
The source picture is a background picture of an animation film called 穢翼のユースティア. The model is reconstructed by only this picture.
Source picture:
Source Picture
Texture
It is only after the reconstruction that we know that the projection of this picture is quite strange. This corridor is much more deep and narrow than what we
have expected. Also the textures also show this point. The shape of the paintings on the following 2 textures are quite abnormal. Also we retouch the textures
of walls to remove the box and the chair.
Texture 1 (Left Wall): Retouched to remove the box
Texture 2 (Right Wall): Be aware of the shape of the painting
Texture 3 (Left Wall): Retouched to remove the box
The VRML Model
These are still images of a new view of the reconstructed scene.
Recap
For the Pentagon, we mainly use the source image 2 to construct the geometric shape of the building. The coordinates are set up by using the box-like building
in the lower-left corner. However, we find that accuracy of the measurement is not quite good. One source of error, which cannot be eliminated by us is that we
cannot make the scale on 3 axis consistent. In source image 2, we assume that the box-like building has a height of 1, width of 12 and length of 8, which may not
be accurate. We can do a better job if we can know the real size of that building.
For that corridor, I think this image doesn't use the same projection as human eyes or cameras do. I think this is usual for paintings because the artists need
to twist the scene sometimes to create the atmosphere they needed. We reconstruct the scene by assuming that the bricks on the floor are square. The result is
that the corridor in the reconstructed scene is narrower and deeper than what it looks like in the original image.