Deep Learning Classification System using PHP and Keras

By: Hossam M. Balaha
C.V.: http://bit.ly/2wAZ2nI

Project Description

This project aims to create an interface between PHP (Web) and Python. The website is developed using HTML, CSS, JavaScript, Bootstrap, jQuery and PHP. You can add a classifier by using the HDF5 exported file from Keras (TensorFlow) and labels file (a text a file where each label in put in a single line). After that, you can use that classifier record to perform different classification. The current progress is performed on images.

Project Tutorial

You can follow the following steps to be able to understand the project.

(1) When you surf the website, you can find the following:

You can change the configurations such as website name, slider info, footer info and cards info from the database tables named: website_configuations, website_cards and website_slider.

(2) You need to create an account. You can register from the Register button in the slider.

You need to enter the required fields: First Name, Username, Email, Password, and Retype Password.

(3) If you created the account successfully, a success message will be displayed and you will be redirected to the login webpage as shown in the following figure.

(4) You need to login with the Username and Password.

(5) You will be redirected to your Dashboard after your success login as shown in the following figure.

(6) You can add a classifier and check your classifiers from the Classifiers card and you can check your classification history from the Classification History card as shown in the following two figures.

(7) You can add a new classifier from the Add New Classifier button in the Classifiers card or from the drop down menu in the navigation bar as shown in the following figure.

(8) You need to fill in the the required fields: Title, Type, Model File and Labels File. The Title field is the nickname of your classifier. The Type field is the type of input data (currently only images are available). The Model File must be the model file with an extension of hdf5. The Labels File must be a text file with an extension of txt. The labels must be ordered in the same order used in the training process. The following figure shows a sample filling of the fields.

(9) The classifier will be added as shown in the following figure.

You can delete it from the delete button (shown in red in the last column: Operations).

(10) You can add a new classification from the Add New Classification button in the Classifications card or from the drop down menu in the navigation bar as shown in the following figure.

(11) You need to select a Classifier and File. For example:

This image is for a patient with a COVID-19. It is just used as an example. The classification process may take a while according to your computer (server) speed and model complexity.

(12) The results are shown in the table in the Classifications History card as shown in the following figure.

The record shows the output of the first three predictions (in percentages with four decimals) in case of you have more than three classes (labels). In our example, it showed that: COVID-19: 99.9968% Viral Pneumonia: 0.0031% NORMAL: 0%

Database Design

The database is designed in SQL using MySQL DBMS. The design is shown in the following figure.

Requirements

• XAMPP for local servers.
• Anaconda with Python 3.
• Create a new environment and upgrade libraries using:
  • Open the command window.
  • Create a new environment using the command: conda create -n tf-gpu tensorflow-gpu.
  • Activate the created environment using the command: activate tf-gpu.
  • Install (and upgrade) the libraries using the command: pip install --upgrade pandas sklearn matplotlib opencv-python.
• Notice that:
  • The name of environment must be tf-gpu as Python will search for that name.

Project Repo. on GitHub

Link: https://github.com/HossamBalaha/Deep-Learning-Classification-System-using-PHP-and-Keras

Copyright

All rights reserved. No part of this publication may be reproduced, distributed or transmitted in any form or by any means, including photocopying, recording or other electronic or mechanical methods, without the prior written permission of me, except in the case of brief quotations embodied in critical reviews and certain other non- commercial uses permitted by copyright law. For permission requests, write to the publisher at the address below. Email: [email protected]