sustainbl.

sustainbl. is an end-to-end food waste management solution that uses modern AI technologies integrated with real-time monitoring for F&B businesses to improve sustainability and make a social impact.

Comment

Inspiration

After going to Georgia Tech dining every day, we inevitably noticed the amount of food that gets thrown in the garbage bin. Additionally, living across a grocery store we noticed the same trend in terms of food wastage. After doing some research we noticed how Georgia Tech pushed for sustainability in the food wastage space in ways such as campus composting, increasing food supply etc. The statistic in this space is very alarming as well and this inspired us to re-imagine what sustainability in food wastage should be.

What it does

  • Real-time inventory management to help grocers with demand expectation.
  • Implements a modular pricing algorithm for perishable items to incentivize purchasing items about to expire to reduce food wastage
  • Recommendation engine for pricing of products, with the ability to find food banks and waste management centers near you to reduce food wastage.
  • Real-time demand projections based on previous sales data using time series
  • Analyzing stock shelves to identify freshness of fruits & vegetables using Computer Vision.
  • Uses tableau to create a dashboard with demand, supply and sales analytics of items in a grocery store

How we built it

We used ReactJS with Chakra UI library to build the front end. For the backend we used LSTMs with tensorflow, OpenCV. APIs used were Google Maps API, Python Flask API, Tableau Visualization API. SQL and firebase were our databases. We also used ExpressJS to connect the frontend with the backend. For data analytics we used tableau. In the backend, we used LSTMs to predict time series demand data. Furthermore, we utilized OpenCV to create a freshness score for vegetables and fruits. We also have a proprietary algorithm that automatically generates prices to increase demand and reduce wastage.

Challenges we ran into

Working with the time series data was difficult because there was a lot of data involved. We ended up using LSTMs - which we are quite proud of. Our frontend was also very challenging, but over the course of the hackathon, we learned and improved. Initially planned to build hardware using a visible light sensor, Arduino, etc to detect whether a particular product is rotten or not. However, due to lack of hardware availability we had to change our approach and achieved similar results with images from a camera using Computer Vision

Accomplishments that we're proud of

We are proud of the computer vision part of our project. It was something out of our comfort zone but we thought it would be useful if you could figure out a way to implement it. We ran into a few challenges when it came to training data and ultimately making a front end for it but it was totally worth it!

What we learned

One of our biggest takeaways was the power of data and cannot only be used to increase consumer behavior by having pricing algorithms but also give valuable insights such as demand, supply and sales of various items to help grocers with demand expectation.

What's next for sustainbl.

  • Expand our user base from grocery stores to restaurants and corporations for optimizing food waste management
  • Implement supply-side algorithms for producers to create a better fit between expected demand and supply.
  • Computer vision algorithm to detect multiple objects at the same time for food freshness detection
  • Connect inventory management APIs to use proprietary algorithms and correlation analysis to provide intelligent insights in order to boost consumer purchasing behavior
Share this project:

Updates