Dinel13 / Go-green-projects

capstone project bangkit 2021

Go Green Project

Garbage Classification and Recyclable Waste Recommendation

For Capstone Project Bangkit 2021

Team Manut || B21-CAP0199

• Cloud Computing Path
  • Huddin ([e-mail]) as Full stack Developer
  • Ainun ([e-mail]) as Cloud Architec and Engineer
• Android Path
  • Alawi ([e-mail]) as Android Developer
  • Dicky ([e-mail]) as Android Developer
• Machine Learning Path
  • Okky ([e-mail]) as Data Scientist
  • Shifa ([e-mail]) as Data Scientist
    

---

Other Repo

• ANDROID REPO https://github.com/dicky7/GoGreen/

---

Other Barnch

• ML API https://github.com/Dinel13/Go-green-projects/tree/ml-API
  • Use Python Flask framework and deploy with Cloud Run on Google Cloud Platform
    
• AUTH API https://github.com/Dinel13/Go-green-projects/tree/backend
  • Use Node.js Express framework and deploy with App Enggine and Cloud SQL as database on Google Cloud Platform
    
• RECOMENDATION API https://github.com/Dinel13/Go-green-projects/tree/recomendation-api
  • Use Node.js Express framework and deploy with Cloud Function and Cloud Bucket as object store on Google Cloud Platform
    
• FEEDBACK API https://github.com/Dinel13/Go-green-projects/tree/feedbac-api
  • Use Node.js Express framework and deploy with Cloud Function Firestore as noSql database on Google Cloud Platform
    
• NEWSLETTER API https://github.com/Dinel13/Go-green-projects/tree/Newsletter-Api
  • Use Node.js Express framework and deploy with Cloud Function Firestore as noSql database on Google Cloud Platform
    
• MARKETPLACE API https://github.com/Dinel13/Go-green-projects/tree/marketplace-api
  • Use Node.js Express framework and deploy with Cloud Run and MongoDB as noSql database
  
  
• FRONT-END https://github.com/Dinel13/Go-green-projects/tree/frontend-web
  • Use React.js, Redux and Tailwind library and deploy with Cloud Run that consume all API service use REST-API from Google Cloud Platform
  • URL https://frontend-rupnuawd4a-et.a.run.app/

---

Design Infrastructure

Backgrounder:

1. 67.8 Million tons waste The Ministry of Environment and Forestry (KLHK) admits that in 2020 the total national waste production has reached 67.8 million tons.
2. Low Recycling Activities Waste management in Indonesia is still in low absorption capacity of recycling activities. This is because Indonesia is still applying the old pattern.
3. 60% Household Waste As much as 60 percent of national waste production comes from household waste. Therefore, there must be good management in the household.
4. Recycling Isn’t Easy Recycling programs vary greatly across the country, and the inconsistency hurts the environment. It’s also confusing and overwhelming for consumers

Machine Learning:

Building models that able to clasificate waste by six label. Build process using baseline experiment, early stopping, checkpoint. Pre-trained model or transfer learning by mobilenetv2. The model was saved with model.h5 and chosen by the [best model] for deployment.

Case :

• Waste Classification
• Recycle Recomendation

Dataset Link:

• Garbage Dataset
  • Trashnet Dataset

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

Waste dataset

Features

• EDA (Exploratory Data Analysis) for Data Tables and Images
• Preprocessing Data and Image
• Image Augmentation
• Callbacks
• EarlyStopping
• ModelCheckpoint
• MobileNetV2

Prerequisites

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

Documentation

1. Download dataset Trashnet Dataset
2. Create the training and validation batch using the train generator.
3. Create the label by using the train generator function
4. Train and validate the model
5. Save model to .h5 file
6. Deploy .h5 model in Flask Rest API
7. Mobile App and Web App consume API with upload image and return the result with JSON

References

• Classification of Trash for Recyclability Status

Thank You :)