Inception Model

Now that we have our input data sorted we can go ahead and implement our network-based classifier model. We will be using a convolutional neural network (CNN) for the classification process. CNN's are neural networks that comprise multiple layers and are generally used to classify optical data.
Convolutional neural networks take the optical input and process them in the model as tensors. Tensors are a representation of the data input in the form of a numbered input multi-dimensional matrix. The optical image taken in generally processed by applications as a 2-dimensional input. Here the model will convert the 2D input into a 4D matrix. ({A Beginner’s Guide t...)
The CNN model we will be using will be Google's Inception-v3, This model has already been pre-trained on all image data from ImageNet. Since this is a pre-trained model the training time required for our classification process will be reduced by many folds and also can be retrained using transfer learning which is the way we can modify the model to our needs. This decrease in training time is possible because we are only training the last layer of the model which will just modify it to be suitable for our application. ImageNet is a database that consists of multiple nodes and each node comprises over 100,000 images which are organized by the WorldNet hierarchy.