DogVsCat-Classifier-CNN

Problem statement : In this Section we are implementing Convolution Neural Network(CNN) Classifier for Classifying dog and cat images. The Total number of images available for training is 8,000 and final testing is done on seperate 2,000 images.

NOTE: YOU CAN DOWNLOAD THE DATASET FROM https://www.superdatascience.com/pages/deep-learning AND CLICK ON THE LINK BELOW Part 2: Convolutional Neural Networks (CNN) Datasets & Templates: <DATASET_LINK_HERE>

Dependencies

Tensorflow 1.10

Python 3.6

Matplotlib

Scikit-Learn

Pandas

Numpy

Install dependencies using conda

Test Train Split

Image training set contain 10,000 images for each category. I split those into 80% train and 20% means test Split each class images into 8,000 for train and 2,000 for test.

Network Parameter:

Rectifier Linear Unit

Adam optimizer

Sigmoid on Final output

Binary CrossEntropy loss

Conclusion

The Architecture and parameter used in this network are capable of producing accuracy of 97.56% on Validation Data which is pretty good. It is possible to Achieve more accuracy on this dataset using deeper network and fine tuning of network parameters for training. You can download this trained model from resource directory and Play with it.