Green Eats 🍽🌿

project description

. /back_end/bin/activate

• Building a web-app that that calculates the carbon footprint of the ingredients in a given recipe and to help users make more sustainable meal choices (ideally recommends more sustainable alternatives.)
• It calculates the carbon footprint of the ingredients in a given recipe and recommends more sustainable alternatives
• The goal of the project is to have an MVP (minimum viable product).
• This means we aim at having something that works, not necessarily the best (in terms of design and range of possible visualizations/functions).

A web application containing:

• A tool to parse ingredients from a recipe and calculate the carbon footprint.
• A system to make personalised sustainable recommendations.
• A dashboard to visualise the carboon footprint of the recipes/ingredients in the database.

Background & Motiavation

• The current worldwide food production accounts for more than a quarter of total greenhouse gas emissions, which in turn is the leading cause of climate change (Ripple et al. 2017).
• Consumers’ food decisions can play a fundamental role in reducing the carbon footprint (Hirsch and Terlau 2015; Verbeke and Vackier 2004) by for example aligning with a more sustainable food diet, where environmentally harmful products (i.e., high in CO2 emissions) have a lower preponderance.
• CO2 emissions associated with food consumption is a relatively new and abstract concept, that consumers cannot (yet) properly digest.
• Therefore, supportive and guidance instruments are needed to help consumers make more sustainable decisions.

Functionality

• Input a recipe by either scraping it from food.com or by manually inputting the ingredients & quantities.
• Estimate the CO2 score per ingredients and per recipe.
• Recommend recipes that could be similar in taste, but with lower carbon footprint.
• Dashboard with meaningful insights about the carbon footprint of the recipes in the database

App architecture

• Database
  • Store the raw recipes scraped by the user
  • Store the parsed recipes
  • Store the CO2 scores corresponding to each recipe/ingredient
  • Store recipe recommendations
  • Store data to be shown in the dasboard
• Backend
  • Connect to the database with the reference data and copy it to your database.
  • Scrape food.com page based on recipe URL.
  • Process & standardise recipe ingredients.
  • Calculate CO2 scores
  • Recommend similar but more sustainable recipes.
  • Write results in your Cloud SQL database
• Frontend
  • Intuitive user interface with interactive tools (input: recipe URL from food.com, output : CO2 score and recommended recipes)
  • Dashboard with relevant insights.

Technology stacks

1. Python (dash, flask, scikit-learn)
2. SQL (postgreSQL)
3. Google Cloud Platform (https://cloud.google.com/)
4. Datastudio (https://datastudio.google.com/)

Focus of deliverables

• We aim at delivering a prototype which can be directly used by researchers.
• Therefore, the focus is the functionality of the web application.
• This means we aim at having something which works well and which can be easily further developed.
• We did not target to develop the best and complex solution with the SOTA AI recommendation algorithm

Development process

1. Set up infrastructure on GCP

• Setup the backend server
• Setup the frontend server
• Setup the database (Cloud SQL server)
• Deploy the demo app

2. Create the app

• Create the back-end APIs (R: plumber; Python: flask /django)
• Create the front-end (R: shiny; Python: flask / django)
• Create the dashboard (Google Datastudio)

3. Deploy and test

• Deploy the backend (first configure to communicate with the database)
• Publish the dashboard
• Deploy the front-end (first configure to communicate with the the backend and the dashboard)