One major problem with computer recognition is that many expect the

computer to see exactly what we see.  But after seeing how a computer sees

objects, it is no wonder why we can’t always get the computer to see what we

want it to see.  One major example is this picture below.  When asked

what this picture consists of, most people would answer a white triangle in the

middle in the picture lying on top of an upside down white triangle with a

black border.  But why do they answer this way?  The most reasonable

explanation is because that is what our visual cortex was taught to see.  Even

though, mathematically speaking, there are no triangles on the drawing.  Most

algorithms are not sophisticated enough to envision what we see.  As far as the

computer is concerned, the picture is of three lines and three semicircles.

    Another problem occurs with optical illusions.  In some situations, the

computer can actually see better then the humans.  Computers cannot fall

victim to illusions that humans do.  Looking at the first picture below,

many people will contest that all the long diagonal lines are not parallel.  But in

actuality, they are, and the computer knows this.  Computers cannot be fooled

by gimmicks, or outside interference to interfere with its perception.  Likewise,

in the second picture, it appears that the center line on top is shorter then the

center line from the bottom.  But like I just stated, the computer knows that

this is false.  So in actuality, the computer could distinguish between ambiguity

above and beyond human capability.