EksFlorasi - Machine Learning

This repository contains source code and documentation for EksFlorasi's machine learning development.

Team Members of C23-PR499

• Machine Learning:
  • M038DSX0335 - Jason Andrew Gunawan
  • M038DSX0340 - James Rafferty Lee
  • M172DSX1831 - Mohammad Azri Harahap
• Cloud Computing:
  • C303DSY0829 - Calista Chandra
  • C016DSX2599 - Sholeh Rodhi Putra Siswantoro
• Mobile Development
  • A038DSX1146 - Frans Wijaya

Machine Learning Development

EksFlorasi is using two machine learning models, to classify images for Flora and Fauna. This section will explain how our team develop EksFlorasi's machine learning models.

Datasets Used

• Flowers Image Classification Dataset
• Animal Image Classification Dataset

Development Environment

We are using Python version 3.9¹ for this project.

Clone or download this repository.

  git clone https://github.com/EksFlorasi/machine-learning.git

Go to the project directory.

  cd machine-learning

Install all required packages.

  pip install -r requirements.txt

Go to notebooks directory.

  cd notebooks

There are four notebooks² of interest:

• modeling/dataset-preparation.ipynb to prepare our datasets
• modeling/fauna-transfer.ipynb to develop a fauna model
• modeling/flowers16-transfer.ipynb to develop a flora model
• inference-test.ipynb to test our models

Machine Learning Deployment

EksFlorasi deploys its models alongside its backend server using JavaScript. This section will explain how we use TensorFlow in JavaScript on our project.

Deployment Environment

We are using Node.js version 16.20³ for this project.

Go to the deployment directory from the root directory.

  cd deployment/image-recognition-js

Install all required packages.

npm install

Run the script after providing your models⁴.

node test-import.js

---

Footnotes

1. Python 3.9: https://www.python.org/downloads/release/python-390/ ↩

2. Each notebook has guides explaining the code and what they do. Should there be an error, it's most likely due to file paths. ↩

3. Node.js 16.20: https://nodejs.org/download/release/v16.20.0/ ↩

4. Due to size limitations, we have uploaded our models on Google Drive. After downloading the models, update the modelPath variable in image-classification.js based off of your model location. ↩