Bangkit-JKT1A

This repository contains the model, and the demonstration of Ocular Disease Intelligent Recognition (ODIR) Classification using Tensorflow as a Final Project of Google Bangkit Machine Learning Academy by team JKT1-A

Team Members:

• Azhari Haris Al Hamdi
• Bimantara Hanumpraja
• Irfani Sakinah
• R. Achmad Dadang Nur Hidayanto

Demo

• Android App (run the installer from Android Directory)
• Web based: ODIR-Classification Site

Slide

Presentation

Repository

Complete project files including model files (h5 format, tflite, and js files) are available to download at Our cloud storage.

Deployment

Android APP

Android package are available to download from Android Directory

You can download the installer and run the application.

Website

To deploy the site, do the following:

First, clone this repository:

$ git clone [email protected]:bangkitjkt/final-project-ODIR5K.git

Run the model to train new data from model directory

$ ipython
$ %run choose_one_model.ipynb

Extract the model into Tensorflow.js files:

$ ipython
>> pip install tensorflowjs
>> import tensorflowjs as tfjs
>> tfjs.converters.save_keras_model(model, 'path_to_tfjs/')

You will get output like:

$ ls -la path_to_tfjs/
total 30420
-rw------- 1 root root 4194304 Jun 20 13:15 group1-shard1of8.bin
-rw------- 1 root root 4194304 Jun 20 13:15 group1-shard2of8.bin
-rw------- 1 root root 4194304 Jun 20 13:15 group1-shard3of8.bin
-rw------- 1 root root 4194304 Jun 20 13:15 group1-shard4of8.bin
-rw------- 1 root root 4194304 Jun 20 13:15 group1-shard5of8.bin
-rw------- 1 root root 4194304 Jun 20 13:15 group1-shard6of8.bin
-rw------- 1 root root 4194304 Jun 20 13:15 group1-shard7of8.bin
-rw------- 1 root root 1767712 Jun 20 13:15 group1-shard8of8.bin
-rw------- 1 root root   21782 Jun 20 13:15 model.json

Then, upload the model to the web server

$ scp -r path_to_tfjs/ remote_username@server_ip:/path_to_model_site

Change model url on index.js file (line 36921)

Last, upload the web to your server

$ scp -r web/ remote_username@server_ip:/path_to_web