af-lazuardi / bangkit-finalproject Goto Github PK

This is our team's repository for completing on the final project about machine learning from Google Bangkit Academy

Our member are:

1. Ida Bagus Agung Nara Surya Darma
2. R. Arif Firdaus Lazuardi
3. Kadek Suar Wibawa
4. Putu Althea Putri Wiradani

1. Jupyter Notebook or Google Colab
2. Python version 3.6 or above
3. Latest version of Tensorflow 2
4. Diabetic Retinopathy dataset

• dataset
• doc
• apps
  • web
  • model

Explanation:

• dataset = dataset that has been downloaded. Local use purpose.
• doc = snaps note and for put all docs about our code
• apps = our apps (both model and client side software)
• apps/web = all of source code for web application
• apps/model = all of jupyter notebook files (save as google colab as notebook) / python files

In this project we build a image classification recognition model that would recognize the digital color fundus photograph from the retinal part of the patient’s eyes, based on the public dataset diabetic retinopathy resized_train_15_19_DG which available on Kaggle. The dataset contain 5 classes which are 0 for No DR, 1 for Mild, 2 for Moderate, 3 for Severe, and 4 for Proliferative DR with total 15 GB sizes and separated into 92.9K image pictures and 5 csv tables.

While develop our model, we used pre-trained VGG without using the top layer from the VGG itself. Instead of using the top layer of VGG, we added AveragePooling2D, two hidden layer, and output layer containing 5 neurons because there are 5 class in this dataset.

As a result from using VGG-19 model, we have got higher validation accuracy. We developed a model that produces 70% training accuracy and 70% validation accuracy.