This is my "faces" Java applet (version of June 26, 2008) for automated
pattern detection. An explanation is below. The algorithm is analysed
in my research paper
Optimising Monte Carlo Search
Strategies for Automated Pattern Detection (pdf)
(J.S. Rosenthal, Far East J. Math. Sci. 32(3), 2009, 311-328).
This applet allows the user to draw an image of their choice, and then
have the computer attempt to find either "faces" (i.e., two eyes and
a nose), or "blobs" (i.e., simple oval shapes) of concentrated dots
on the image. An explanation of the image controls (yellow buttons)
is as follows:
The user may also click on the image itself to draw
image dots (one dot at a time).
- SPRINKLE DOTS: place 100 image dots uniformly at random
- SEARCH ONCE: perform a Monte Carlo search for a blob,
clearing any previous searches
- SEARCH AGAIN: perform a search for a blob, without
clearing previous searches
- CLEAR SEARCH: erase any previous searches
- DUMP DATA: dump the current image data to the Java
virtual console (for testing purposes)
- RESTORE DATA: read in a particular test image,
which consists of a simple face-like image near the center, plus a few
random dots. (More random dots can then be added by clicking on
- BLOB / FACE: select whether to search for simple
blobs (ovals), or faces (two eyes and a nose)
In addition, the applet's control buttons (blue) allow for adjusting a
number of parameters of the search algorithm. For more details, see my
recent research paper (or
images). See also the source code.
Note that the algorithm does not "cheat" by making use of the order in
which the dots were created, nor which dots were created by the user and
which ones at random. Rather, it makes use of the final image array
and nothing more.
In the future, I hope to allow for importing external images
(e.g. photographs) rather than just drawing them.
My interest in computer vision was inspired by a talk by Yali Amit and a meeting
with Sven Dickinson.
[Jeffrey Rosenthal /
contact me /
Applets Page /
For archiving purposes,
my previous applet versions are also available: