Dogs-vs-Cats-Redux-with-CNN
This is a little explanation so I will be a little informal to talk about this work
This is my first attempt at Kaggle and building a CNN from scratch. I hope you enjoy
First I want to apollogize for my few mispelling errors.
Second I want to say that are a few differences between the .py and jupyter file.
You can check my original file (at this moment in version 3) at Kaggle Kernel: https://www.kaggle.com/rezende/my-first-cnn-network-fom-scratch
I was mixing between them but I decide to finish it in Kaggle's Kernel running in 100 epochs which means a difference of 60 epochs from my .py file because it is little bit fast and do not stress so much my equipment since I was running it on a notebook with a GTX 1050. But in case you run with only 40 epochs the result was "acc: 95.38799999999999%". Pretty close to 100 epochs which was 96.576%.
The .py file is more suitted to use in terminal with some comments and best visual aesthetic I could make for a terminal (like if it is possible)
In terms of building a CNN everything was done by what I learn by watching DeepLizard videos and read in Deep Learning with Keras by Antonio Gulli, Sujit Pal.
To pre processing my data I want to give a special thanks to sentdex and RMOTR.
Well. That is it. I hope you enjoy this model