docbot-bangkit-2021/docbot-machinelearningmodels

This repository is used to create Machine Learning models. Building three kinds of models that include covid detection, fruit and vegetable nutrition content, and general disease detection.

DocBot

Profile

Team ID : B21-CAP0362

Members

• M2082044 - Nur Imam Masri
• M2242154 - Rafkah Rizqia
• A0131304 - Dian Arisma Wicakso
• A0141345 - I Made Ari Widiarsana
• C2242155 - Nurul Uswatun Hasanah

Roles/personnel

• Project Plan (M2082044 - Nur Imam Masri)
• UI/UX (C2242155 - Nurul Uswatun Hasanah, M2242154 - Rafkah Rizqia)
• Build Machine Learning Model (M2082044 - Nur Imam Masri, M2242154 - Rafkah Rizqia)
• Android Development (A0131304 - Dian Arisma Wicakso, A0141345 - I Made Ari Widiarsana)
• Deployment Application (C2242155 - Nurul Uswatun Hasanah)

DocBot-ML

This project is our final project for Google Bangkit Academy 2021.

Android: DocBotApp

Cloud: DocBot-Cloud

Backgrounder:

The development of the health sector in Indonesia is still not growing optimally. Public submission of symptoms of disease requires early solutions in dealing with emerging diseases. As well as long-distance service in some areas is still very difficult. The government established an institution, namely Puskesmas, which is a health service facility. Public health problems, especially remote reports, recommendations for early treatment, and appropriate health programs for the community, are still lacking and not yet integrated. The strategy in the form of application development by utilizing deep learning aims to analyze the symptoms of the disease from the community in conducting classification based on disease diagnosis health test data. It is hoped that it can help people who need recommendations for first action against symptoms of disease and related health information, in addition to that expected from the data. which is received can be analyzed by the Puskesmas to design a health program.

Machine Learning:

Building three kinds of models that include covid detection, fruit and vegetable nutrition content, and general disease detection. Build process using baseline experiment, early stopping, checkpoint. Pre-trained model or transfer learning by resnet152v2, densenet121, inceptionv3, mobilenetv2, vgg19. The model was saved with model.tflite and chosen by the best model for deployment.

Case :

• Covid19 Classification
• Skin and Eye Diseases Classification
• Fruits and Vegetable Classification

Dataset Link:

• Covid19

  • Early stage symptoms of COVID-19 patient's

• Fruits and Vegetables → Combined Dataset FruitsVegetables : Github | Drive | Kaggle

  • Crapping Image on Google Search (Manually)
  • Fruits-262
  • Fresh and Stale Images of Fruits and Vegetables
  • Tomato Cultivars
  • Kaggle Fruits 360 dataset
  • Fruits & Vegetables
  • fruits_vegetables_photos

• Disease → Combined Dataset Disease : Github | Drive | Kaggle

  • Crapping Image on Google Search (Manually)
  • Dermnet
  • Various Skin Diseases Dataset

Preview of the image and data used are shown in the picture below.

Early stage symptoms of COVID-19 patient's

Eye and Skin Disease

Fruits and Vegetables

Open In Colab

Covid-19

• Covid19_Baseline_Model_01.ipynb

• Covid19_Baseline_Model_02.ipynb

• Covid19_Baseline_Model_03.ipynb

• Covid19_Baseline_Model_04.ipynb

Disease

• Disease_Baseline_Model_01.ipynb
• Disease_Baseline_Model_02.ipynb
• Disease_Baseline_Model_03.ipynb

FruitsVegetables

• FruitsVegetables_Baseline_Model_01.ipynb
• FruitsVegetables_Baseline_Model_02.ipynb
• FruitsVegetables_Baseline_Model_03.ipynb

Features

• EDA (Exploratory Data Analysis) for Data Tables and Images
• Preprocessing Data and Image
• Image Augmentation
• Callbacks
• EarlyStopping
• ModelCheckpoint
• ResNet152V2
• DenseNet121
• InceptionV3
• MobileNetV2
• VGG19
• TFLite x labels

Prerequisites

1. Jupyter Notebook or Google Colab
2. Kaggle API Token → Generate
3. Python version 3.6 or above
4. Latest version of Tensorflow 2.5 (or you can update again by rerunning .ipynb and updating models)

How to use

1. Create kaggle.json from Kaggle
2. Go to your Kaggle profile then download your Kaggle API.
   • My Account → Look for API section → Create New API Token
3. Open .ipynb with Google Colaboratory using open in colab
4. "Save a copy in Drive" to run and edit with your account. Click File > Save a copy in Drive in your Google Colaboratory.
5. Upload your kaggle.json if asked to upload it.
6. Will download the dataset on kaggle
7. If using GoogleDrive for the dataset, GoogleAuth click on the given link and sign in with your Google Account.
8. Done :)

References

• http://bit.ly/papercapstone01
• http://bit.ly/papercapstone02
• http://bit.ly/papercapstone03
• http://bit.ly/papercapstone04
• http://bit.ly/papercapstone05
• http://bit.ly/papercapstone06

Thank You :)