Finding patterns

A very simple approach to identifying a given image containing the number 0 (or not) can be solved by storing images of all the numbers and comparing every image in order to identify a good match. This will be a tricky and tedious process, since computers do math literally. Until we have an exact match of the input image with the one in our repository, we will not find a match. From a computer's perspective, an image will be seen as a two-dimensional array of pixels with a number at every position. Let's see an example of this with the number 0:

The diagram on the left within the preceding diagram is in our repository, while the one on the right is the input image, which is a slightly deformed/hand written version of number 0. Upon verification, the computer will try to match all pixel values. However, when there is no exact value match at the pixel level, the number 0 won't be identified. This is where we need a CNN to help.