Cat-Vs-Dog-with-Data-Augmentation

This Repository help user to understand Data Augmentation Concept on Cat and Dog dataset which is small dataset.

Data Aumentation is data analysis techniques used to increase the amount of data by adding slight midified copies of already existing data.

Data Augmentation act as a regularizer and help to reduce overfitting when training a Machine Learning model.

Data Augmentation is helpful if you have small data.

Streamlit Online Application

Where you can upload the image to check whether it Cat or Dog using pre-trained model.

Streamlit Appilaction

How to Install Dependencies to Run the application.

The following bash code will install all the necessary dependencies.

git clone https://github.com/Arpitkamal/Cat-Vs-Dog-with-Data-Augmentation.git
cd Cat-Vs-Dog-with-Data-Augmentation
pip install -r requirements.txt

how to run in local machine

To run the application go the downloaded Directory and run this command

streamlit run main.py

To run the application where you can test the pre-trained model by uploading a image.

streamlit run application.py

Stremlit Application

In this Application user can Train the Convolutional Neural Network with or without Data Augmentation.

User can Build the Model by Selecting the Hyperparameter

By Selecting following Hyerparameter:

Activation function for the convolution Neural Network.
Padding for the Convolutional Neural Network.
Filter size for the First Convolution Layer.
Filter size for the Second Convolution Layer.
Filter size for the Third Convolution Layer.
Filter size for the Fourth Convolution Layer.
Activation function for the Dense Layer.
Number of Epochs you want to Run the Model.
loss Function for model Compiler.
optimization Function for the model Compiler.

User can select Different Parameters for Data Augmentaion

By selecting following Parameters:

Rotation range is a value in degrees (0–180), a range within which to randomly rotate pictures.
width shift range and height shift range are ranges (as a fraction of total width or height) within which to randomly translate pictures vertically or horizontally.
Shear range is for randomly applying shearing transformations.
Zoom range is for randomly zooming inside pictures.
Horizontel Flip is for randomly flipping half of the images horizontally. This is relevant when there are no assumptions of horizontal assymmetry (e.g. real-world pictures).
File mode is the strategy used for filling in newly created pixels, which can appear after a rotation or a width/height shift.

Once user selected all the above parameters. User can click the Train Model button to train the model.

After clicking Button application show the Model Summary and start training the model.

In the end Application will Display two plot:

1. Training and Validation Accuracy
2. Training and Validation Loss

Other option

If you don't want to run Streamlit application just clone the repository and run testing.ipynb file to see the work of Data augmentation.