CNN

CNN is a deep learning model that able to extract image features within the training process.

Why CNN is able to do that? It was something related to computer vision.

feature extraction of image

Image's features are points, edges and lines.

By walking through the image with a special mask, we can have score for each pixel which albe to
represent this pixel is on line or edge, etc.

In CNN, the mask and extraction process were inside Convolution Layer.

Another layer I would like to introduce is: Pooling Layer

pooling layer are used to downsampling.

While image data were affect by lights(Intensity variations), size of origin image(),
by only keep the maximum value of the sliding window, we can get the most importance part in mask and avoid these affects.
Also progressively reduce the amount of parameters. :)