Explanation

SkyAfra is a mobile app to classify an image whether it is a biodegradable or non biodegradable food. In this project, we decided to make a waste segregation classification that will classify an image to 2 labels: Biodegradable and Non-Biodegradable. We used MobileNet V3 for our base model to make sure the model is accurate, fast, and lightweight since we want to deploy it and connect it with our Mobile App.

Project Goals

The project aims to develop a machine learning model using MobileNetV3 that can classify images to 2 labels: Biodegradable and Non-Biodegradable. We also aim to deploy the model on the cloud and integrate it with the mobile app.

Tech Stack

We used various technologies to support us developing the application. For mobile development we used Kotlin. We also used Tensorflow to create the machine learning model. And then we developed the backend using Node.js, and we used Google Cloud to deploy our model.

Features

This project's main feature is image classification, get yourself an image and SkyAfra will classify it.

My Contributions

As a machine learning engineer, I am responsible for gathering data, cleaning data, preprocessing data, and creating the model. We gathered the data from Kaggle. The machine learning model is created using MobileNetV3 on Google Colaboratory. I also helped Cloud Engineers team to deploy the model.

Problems

The main feature we wanted to build was actually real-time classification, but we considered it will cost too much time. After had a discussion with our mentor, we decided to create an image classification instead. We didn't get enough data to make our model good. We got the accuracy of 86% but it was overfitting. Then we did a hyperparameter tuning and used callback functions to prevent overfitting.

Shoutout

Shoutout to my teammates for spending their time to finish this project together.

1. Andreas Fernandez W (Machine Learning)
2. Saffanah Nur Fadilla (Machine Learning)
3. Revanantyo Dwigantara (Machine Learning)
4. Ken Anargya Alkausar (Cloud Computing)
5. Adrian Ismu Arifianto (Cloud Computing)
6. Yopira Eka Putra (Mobile Development)
7. Fardilla Bara Hidayah (Mobile Development)