mouselearning

An AI Diet Mobile Application which aims to helps users lose weight in order to avoid health problems, such as heart-related issues and asthma.

Architecture

The application consists of three main parts: front end (client), back end API, and the machine learning application. Each of these three services communicate between each other, sending and receiving requests for each of the application's features. The back end API and the machine learning application are containerised using Docker. The Darknet YOLO Model used for the food detector in the machine learning application has been set up with Tensorflow Serving to future-proof the application by easing the process of using a new or enhanced model in a production setting without changing any of the service architecture or APIs.

To find out more about deploying deep learning models using Tensorflow Serving, Docker, and Flask, click here.

To find out more about training a YOLO Food Detection Model using Darknet, click here.

Team Members

1. Nardiena Althafia Pratama - 46223713
2. Muhammad Naufal - 46223638
3. Davin Iddo Irawan Alfian - 46223740
4. Owen Jordan - 45802942
5. Huu Minh Quan Tran - 45262612
6. Aditya S Hadinata - 43642498

Option 1: Set Up Without Docker Compose

Prerequisites

1. Docker
2. Python 3.6.0 and 3.7.0
3. Tensorflow serving - it is recommended to set this up using docker
4. React Native and its development environment - choose the React Native CLI Quickstart option in this link

Environment Set Up

Two python environments are needed to run this project, one with Python 3.6.0 and one with Python 3.7.0.

First Environment: ML Flask Server (Python 3.6.0)

1. Run cd plateducate-ml/flask_server and run pip install -r darkflow.txt.

If your system does not have a dedicated GPU, replace the tensorflow-GPU module with tensorflow. Otherwise, leave the file as it is.

3. Enter the darkflow directory by running cd /darkflow.
4. Run python setup.py build_ext --inplace and pip install . to install the darkflow open source library.

Second Environment: BE Flask Server (Python 3.7.0)

1. Run cd plateducate-be/ and run pip install -r requirements.txt.

Database Set Up

1. Install MySQL here.
2. Export the schema using the .sql file included in the repository.

Execution

Machine Learning Application

1. Tensorflow Serving

On one terminal, do this step:

1. Run tensorflow serving by running this command in your terminal:

docker run -p 8501:8501 --name=object-detection --mount type=bind,source=PATH-TO-PROJECT/plateducate-ml/serving/conf/,target=/tensorflow-serving/conf/ --mount type=bind,source=PATH-TO-PROJECT/plateducate-ml/serving/model-data/,target=/tensorflow-serving/models/ -t tensorflow/serving:1.13.1 --model_config_file=/tensorflow-serving/conf/tensorflow-serving.conf --model_config_file_poll_wait_seconds=60

2. Machine Learning Server

Open a different terminal and do these steps:

1. Run cd/plateducate-ml/flask_server/
2. Run flask run.

Back End Server

Open another terminal and do these steps:

1. Run cd/plateducate-be/.
2. Run flask run -p 4000.
3. Ensure that the URLs in the backend endpoints refer to localhost instead of the Docker container name, since we are not using Docker Compose.

Front End Client

Open a terminal and do these steps:

1. Run cd plateducateFE/.
2. Run npm i to install the needed modules.
3. Run npx react-native start.

On a different terminal, do these:

1. Run cd plateducateFE/.
2. Run npx react-native android-start.

The application should now start running on Android Studio.

Option 2: Set Up With Docker Compose (Linux and GPU Only)

Platform and Hardware Requirements

1. Linux Distribution from this list
2. NVIDIA GPU

Prerequisites

1. Docker and Docker Compose
2. NVIDIA Container Toolkit
3. Tensorflow serving - it is recommended to set this up using docker
4. React Native and its development environment - choose the React Native CLI Quickstart option in this link

Database Set Up

1. Install MySQL here.
2. Export the schema using the .sql file included in the repository.

Execution

Machine Learning Application

On one terminal, do the following:

1. Enter the plateducate-ml directory by running cd plateducate-ml/.
2. Run sudo docker-compose up --build.

Backend Server

Open another terminal and do the following:

1. Enter the Plateducate project directory (highest level).
2. Run sudo docker-compose up --build.

Front End (Client)

Open a terminal and do these steps:

1. Run cd plateducateFE/.
2. Run npm i to install the needed modules.
3. Run npx react-native start.

On a different terminal, do these:

1. Run cd plateducateFE/.
2. Run npx react-native android-start.

The application should now start running on Android Studio.