Interesting Examples
-
One-dimensional (Gray-scale Features) Model
-
Two-dimensional (Color Features) Model
-
Two-dimensional (Motion Features) Model
One-dimensional (Gray-scale Features) Model
1.1 Sheep Dog (FarmVille Game)
Source:
Video:
1.2 Elephant (FarmVille Game)
Source:
Video:
Two-dimensional (Color features) Model
2.1 White Tiger
Source:
Results:
Video:
2.1 Lion In Black & White
Source:
Results:
Two-dimensional (Motion features) Model
3.1 Highway Traffic Camera
Source:
Video from Youtube. [>]
Frames Taken:
Results:
Comparison with Color based model:
|
|
 
Motion Based |
3.2 Rolling Ping-Pong Ball
Source:
Recorded using my Phone Camera.
Frames Taken:
Results:
Comparison with Color based model:
 
Color Based |
Motion Based |
3.3 Rolling Red Stress Ball
Source:
Captured using my Phone Camera.
Comparing Results:
Color Based |
Motion Based |
Why Motion Based Segmentation Fails Here?
Let's take a look at the feature image, which in our case, is the Optical Flow of the two rolling red ball frames. Below are each of the dimentions:
 Velocity over X-Axis |
 Velocity over Y-Axis |
"Everything that we see is a shadow cast by that which we do not see"
The moving shadow below the red ball misled the Optical Flow estimation, which subsequently led the segmentation algorithm astray.
3.4 Moving Cat
Source:
Result:
