The Carbon Tracker

One of the prominent problems that arose during the creation of our project was the scenario in which the product's barcode had been removed or if the food item had been prepared as opposed to being purchased. The solution we came up with is using a neural net to predict what the food is. We trained a ssd_mobilenet_v2_coco model using Tensorflow's Object Detection API. The reason we chose to use a mobilenet is that we hope to eventually make this usable on a mobile device meaning the model must be small and quick. Therefore an ssd_mobilenet was the best option with a guess speed of 27ms with minimal loss to the accuracy. The model was trained on a Nvidia 1050 Ti GPU for 11 hours, with over 100 images for each food.

HOW IT WORKS: Take a picture of the food using your webcam or phone camera, upload it to the Server which will call our food model API to make a prediction about the food, and return carbon footprint data about the food so you can make the best decision for the environment.

In the event where the barcode on the product is present, we created a web scraper to fetch information on a UPC database about the product. The scraper was programmed using Python 3 and it implemented beautiful soup for data collection. After intense and thorough research, our team created a Python dictionary listing the carbon footprint of various categories of food.

HOW IT WORKS: When you run the program, all you are required to do is type in the 12 digit barcode located on your food product. Then, the program will fetch the product from the database and display the estimated carbon emission level (carbon footprint) produced during the creation of the product.

• Divide and Conquer
• Front End, Design, and Server - Ashwin Babu
• Vision Processing and Neural Network Training - Alvin Zheng
• Barcode Scraper Programming - Rishi Peddakama
• Data Analysis and Modeling - Raadwan Masum
• Collaboration Resources
• Git - Constantly syncing changes to Github repo so that everyone is on the latest code
• VSCode Live Share - Add-on to source-code editor, Visual Studio Code which allowed efficient updating and collaboration through the programming process

The purpose of our product is to spread awareness of our changing environment and emphasize the importance of changing now before its too late. Granted, it is very hard for the average working class citizen to find time to help preserve the environments through local cleanups and events. And that is why we created our website; It is a simple carbon footprint tracker that tells you how much C02 is being released with any food product you may own that can be identified through your webcam or the barcode on the product. We hope that with this, people are able to change the environment one little step at a time through the surveillance of personal carbon emissions.