Use CNN with Keras to classify Dog & Cat Images
Goal:
- To succesfully apply the concept of Convolutional Neural Network to large scale data.
- Build a classifier which successfully differentiates Cats and Dogs
Data:
Dogs vs Cats dataset from Microsoft is used to train and build a convolutional neural network model to distuingish cats from dogs. The dataset consists around 25000 images equally divided into cats and dogs.
Approach:
First I downloaded the dataset to my local and then uploaded it to my Google Drive so that I can use it in Google Colab. Afterwars I prepare the data by purposely reducing the quality of each image for 2 reasons one there may not be such high quality images to classify and other being reduction in comptational cost such that the images can still be recognized to some extent
Results:
After 10 or so iteration 89% train accuracy and 88% test accuracy is not bad but ceratinly can be improved. I believe the Classifier performance can be certainly improved, in future iterations I'll try to further reduce bias and variance.
Citation:
- This is a rework of Convolutional Neural Networks - Deep Learning basics with Python, TensorFlow and Keras work from Python Programming Tutorial
- This analysis or model building was done using Google Colab