Classifying number images using NN and CNN

Description

This is a project that compares the image classfication performance of two different neural networks:

• A conventional fully connected neural network (NN)
• A Convolutional Neural Network (CNN)

Results

Conventional neural networks are limited by its own nature when it takes to analyze images, that is why they can classify images but not as efficient as a CNN can. CNN are able to get patterns and difference one class of the others. Conventional NN are able to classify but can not obtain any patterns because they base their knowledge in the analysis of the whole image preventing it to get any pattern. In my analysis, both types of neural networks are able to classify quite well, but CNN is capable of obtaining better accuracy with less cost.

Neural Network

• Accuracy:
  
• Loss:
  

Convolutional Neural Network

• Accuracy:
  
• Loss:
  

Other data

• Author: Samuel Arrocha Quevedo