Cat vs. Dog Image Classifier

PROJECT OVERVIEW

Image classifier trained to distinct between cats and dogs images. Convolutional Neural Network was built with Keras & Tensorflow 2.0(GPU). Heroku-hosted web application was built with Flask framework.

Kaggle Dataset

CONVOLUTIONAL NEURAL NETWORK CHARACTERISTICS

1. Image Input Shape - 128,128,3, activation - relu
2. Transferred Learning from Google MobileNetV2 with customized head
3. Compiler - optimizer = 'RMSprop', loss = 'binary_crossentropy', metrics = ['accuracy']
4. Acc - 92% Loss - 19% (approx 30min/epoch on GPU)
5. CNN Code Location: deep_learning/ConvolutionalNeuralNetwork_V2.py

TOOLS, MODULES & TECHNIQUES:

Travis CI

Python – web development:

Flask

Python – CNN:

keras | tensorflow | scikit-image | pandas | numpy | h5py

Web Development:

HTML | CSS | Bootstrap | Materialize

Thank you,

Lukasz Malucha