Convolution Neural Network(CNN)

The subject of machine learning that deals with the prediction of images and (visuals data). these topic falls under the category of CNN.

Images are the collection of independent units called pixels. And each pixels represents one feature, so as well all know the normal photo are of 500 X 500 pixels, so the challenge here is for the calculation of weight for each features(pixels here) it’s hetic job, hence what we need to do is to reduce the dimension of the image or reduce the less important feature

as we know that dimension is equivalent to the number of features(Here, pixels ), so reducing feature means the reduction of Dimension (but one thing should keep in mind regarding the data loss)

the above-mentioned condition can be achieved with the help of Convolution and polling.

Convolution is an algorithm for achieving the feature-learning specially the important feature, which is done with the help of edge detection. This feature selection achieved by Convolution layer.

CONVOLUTION LAYER

This layer consist of feature map(group of filters/kenals) and these kernals/features helps in the detection of edges, further edges are the one that helps in important feature selection . we also need to add the number of filters.

note:- we take the feature map, as here the Keras module will pick the filters which will best fit, in that particular case

Pooling Layer

Polling is another layer that helps in reduction of dimensions (it like converting 4 pixels into one for the less important part of images)

NOTE:- Its is not necessary that we need to convolve and pool single time , multiple layer can be introduced

Flatten Layer

Once we are done with the C part of CNN , we need to create the NN dense layers. so to feed the data inside the NN, what we need is ,the data should be in 1D format , for this formatting we need another layer called Flatten from Keras module