2013-05-08

Perceptra 2D Test Run – 1 8th May 2013

There is a dramatic difference between these two approaches. The first one creates all the rectangles including the ones from the borders to the edge of the digit. The second one only creates the rectangles that make up the figure (do not extend to the left or right borders). This means there is valuable information in the rectangles that have a border. Since the border is always the same the information must be the edges of the figures. Has my approach of building shapes out of rectangles been not as beneficial as creating compositions of edges? That's what I need to try next. In a linear 1D array of sensors shapes are effectively two edges separated by a distance. Isn't this a composition of two edges? If I was to just combine edges and then combine pairs of edges etc. would I not come up with the same result?