Computer Vision - Project 3

Galen Wilkerson

 

Artifact 1:

Fallingwater, by architect Frank Lloyd Wright

 

Image with model overlay:

 

 

These are two texture maps from the image:

This is the triangular one in the bottom left of the image.

It was interesting to find the left-most point, since I had to guess that it was below the balcony.

 

This one of the window adds some color to the final vrml, since most of the house is white or grey:

 

Rotated vrml screenshots:  (click here to see the vrml)

 

 

    Methods:

The front left corner (nearest point) served as my reference plane, and I was able to work my way horizontally and vertically using "same z" and "same xy".

It was a puzzle of how to work my way up to the white balcony above until I noticed that the awning was just above the right-hand corner of the lower balcony. (far right)

As an experiment, I also tried to guess at the height of branches in the foreground, but it ended up looking strange, so I discarded them.

The right-hand chimney is still a little wrong, but I've OD'd on polygons lately.

 


 

Artifact 2: 

Edvard Munch "The scream"  (Der Schrei)

 

 

Image with model overlay:

 

 

Two texture maps from the image:

 

This is the "land" beyond the railing.  I modified the right side of this, since it was obscured by the figure screaming and the railing.

 

This is the right-hand figure in the background.  They are kind of menacing, aren't they?

To make cutouts of the figures, I had to set the point at the top of the head to be sameXY as the feet,

then use a straight vertical line to find sameZ coords of the points on the shoulders, etc.

 

 

Rotated vrml shot: (click here to see the vrml)

 

 

Another one: "Screaming on Jupiter"

 

 

It was difficult to cutout the figures and to create polygons that were obscured by other polygons. 

Also, the sky in the background presented some challenges, since it and the hills were too far away, and the homography kept giving weird values for it's x and y coordinates. 

Finally, I just gave one corner some coordinates and computed everything relative to that, resulting in the jovian background you see here.

 

Soon I can sleep...

 


 

For fun, I started to make an extra artifact from a photo I found on National Geographic's website:

 

 

Here's the model in progress:

 

 

 

 

 

(click here to see vrml)