Image Classification App

Overview

• This project is a machine learning project that builds a Cinder image classification app from scratch
• Algorithm: CNN (Convolutional Neural Network)

Setup

• libraries:
  • Cinder
  • Eigen 3.4.0
  • stb
• note:
  • this project root directory should be placed in a subfoler of the cinder library
  • set the working directory of the app to the root directory of the repo to avoid file IO path issue

How to use the App

Main Page

Model Training and Saving

• Press c to configure params before dropping the dataset folder
  
• Drop an image dataset folder to start training. Dataset must follow structure specified in util.h. It may take a while to train the data depending on dataset size. A training analysis graph will pop up on complete.
  
• output .cnn model file will be saved to \data folder

Uploading Model File

• Model must have extension of .cnn with correct format
  

Make A Prediction

• A model must be present before making a prediction
  
• Simply drop an image file into the app window