RetrievR-Sounds like a big yellow dog!

I've been intrigued by RetrieveR. I made a few scribbles and now I'm trying a systematic approach to determine how this tool works. This is the one in which you can draw a little something and similar images from Flicker will be found. Obviously color is the first filter, but after that, what's going on? With apologies to Albers, Rothko and Newman, I colored the bottom half of the rectangle hot pink and then switched the composition-pink on top and white on the bottom. Next I colored a flamingo square in the middle of a white ground and again, I reversed it. Finally I used black to draw on top of the colored areas. From my "research" there are three filters or indexes--color, placement and shape which includes line.

Some retrieved images leave the viewer wondering, "Why in the world..." With other retrieved images it's possible to see that placement matters. They contain an area on the side or top or bottom that matches the drawn image. The addition of lines on top of the colored areas appears to read more as a change of color and as a shape in the retrieved photographs. I am left wondering whether there is a strict order in which the attributes of the drawn image are evaluated or if the sequence of evaluation changes. I question, too, whether the images are evaluated from a few select spots or from the entire image. A fairly amusing half hour, although, I can't yet envision a use for this tool.