{\rtf1\ansi\ansicpg1252\cocoartf1404\cocoasubrtf470 {\fonttbl\f0\fswiss\fcharset0 Helvetica;} {\colortbl;\red255\green255\blue255;} \margl1440\margr1440\vieww10800\viewh8400\viewkind0 \pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural\partightenfactor0

\f0\fs24 \cf0 \

Train-Test-data/ contain two matlab files converting .png files into one csv files for training and testing set. Please download the Format1 file from the website:

http://ufldl.stanford.edu/housenumbers/

Tensorflow-cnn-SVHN-4-4layers-Dropout-40/

contains convolution neural network of 4 layers and drop out applies to each layer but
not the input layer. The pictures are of three channels and 40 by 40 pixels.

Tensorflow-cnn-SVHN-4layers-DropoutMaxout54grayscale/

contains convolution neural network of 4 layers and drop out applies to each layer but not the
input layer. The picture are of 54 by 54 pixels.

The svhn.py file will do the training and testing for you on training and testing set, after training, the model.ckpt file will be generated.
Online-testing/ help with the real short number sequence recogintion.

Shape the png file to 40 times 40, 3 channels and copy it into the same folder. Open the RestorePNG.py.

In the main function, such as def main(): img=mpimg.imread('28.png');

Put the name of the png file that you want to do the recognition. Run the RestorePNG.py.

}# Project-OCR