cat-dog-cnn-classifier

Description

The dataset used on this classification model comes from a competition that aimed to develop an image classifier trained from images with dogs and cats. This dataset can be accessed clicking in the following link: Kaggle Cats and Dogs Dataset

Features

The attributes on this dataset are the information contained on every single image as an array of pixels [Black:0 | White:255]. Every array has the following shape: [image_width, image_height, channel]. Originally, the images contain 3 channels, one channel for every color (RGB).

Classes

• class 1 : dog
• class 2 : cat

Dependencies

You can use pip or conda to install the dependencies:

• tensorflow
• matplotlib
• jupyter
• pandas
• pillow
• scikit-learn
• imageio
• OpenCV

Usage

If you want to try this program, download this repo and launch jupyter to run it on your machine.

- - - TODO - - -

• ENVIRONMENT PREPARATION

  • Install library dependencies
  • Document installation and usage

• DATA EXPLORATION

  • Add dataset description
  • Preview the structure of the dataset
  • Add data visualizations

• DATA PREPROCESSING

  • Apply standarization to feature data
  • Apply one-hot encoding to categorical data
  • Split data into training and testing sets
  • Output preprocessed data for faster preloading

• DATA ANALYSIS

  • Define network parameters
  • Define network structure
  • Add different network configurations
    • Define learning rate with different decaying methods
    • Set up cost, optimizer, and accuracy function with different configurations
  • Define model execution
  • Visualize evolution of MSE on training and testing datasets through epoch iteration
  • Visualize evolution of loss function
  • Visualize evolution of learning rate
  • Add log and summary writer
  • Add Tensorboard visualization
  • Add checkpoints for model restoration

• MODEL DEPLOYMENT

  • Load a pretrained model
  • Test it with new data

• OTHERS

  • Update README files
  • Update all nbviewer links
  • Add Tensorflow 1.x, Tensorflow 2.x, keras, tf.keras, and scikit-learn data analysis notebooks