iman-sharifi-ghb/CNN-Street-View-House-Number-Detection-SVHN

We use Convelutional Neural Networks (CNN), as a Deep Learning paradigm, to detect the house numbers using the MNIST dataset.

Street-View-House-Number-Detection-Using-CNN

In this project, we aim to develop a Convelutional Neural Netwrk platform, as a Deep Learning algorithm, to detect the house numbers using the MNIST dataset. We could achieve 95% accuracy in the training process and also compared our algorithm with a Fully Connected Layer, placing an emphasis on the reliabality of the network.

CNN Structure:

(3*3) Kernel Models:

(5*5) Kernel Models:

CNN Results with Kernel Sizes:

Results shows CNN Model detected House Numbers with high precision.

Fully Connected Models:

Fully Connected Results:

Precision of This Method is lower.