i am working in the field of image processsing and we have a certain problem which is difficult to solve by rules-defined algorithms. Is it possible to teach a certain pattern into a neural network, and search for this pattern in a larger image?
For example ,the pattern is 20x20 pixels, but the complete image is for example 2000x1000 pixels? The size of the pattern should also be flexible, the algorithm should also find patterns which are slightly larger or smaller then this predefined 20x20 - in short - the size of the pattern should be flexible.
Is it possible to teach this pattern to a neural network as an isolated input? This means, you teach the 20x20 cropped image, instead of the whole 2000x1000 pixels image?
I hope there is anybody out there who can help me, i know it is a bit difficult to describe.