EcoNibbles

Inspiration

What it does

Inspiration

An inspiring factor that motivated us to pursue this topic was the growing awareness of the critical importance of sustainability and environmentally responsible practices today. While there are numerous tools for assessing the nutritional content of food, we couldn't help but notice that there is a lack of awareness of the environmental impact of the food we consume. This realization served as a powerful catalyst for our project. We were driven by a desire to bridge this gap and provide individuals with a comprehensive tool that not only evaluates the nutritional value of their meals but also quantifies the environmental footprint associated with their dietary choices.

What it does

By simply inputting an image to EcoNibbles, EcoNibbles is able to generate the environmental impact of your meal. Our app analyzes the image that you input, then calculates the carbon impact of your meal based on the individual ingredients it is made of. EcoNibbles allows consumers to take individual action in humanity's fight against global warming. Moreover, this app empowers caring people like you to make a difference.

How we built it

We utilized a database with existing pictures of foods and dishes and their corresponding ingredients. We then fed the images to the machine learning model which classifies the type of meal, and then extracts a list of ingredients from each image. Then, we used a web scraping and OpenFoodFacts API to collect the CO2 data for each ingredient. The entire application was then wrapped in a web based application, complete with a front and backend.

Challenges we ran into

Throughout this endeavor, we encountered challenges that tested our technical and problem-solving abilities. Fine-tuning the machine learning model to achieve our desired results presented a set of complexities. With so many hyperparameters to tweak and elements of the neural network, getting the machine learning model working was very challenging.We also had to address issues arising from a dataset website that proved prone to occasional crashes due to the high volume of API requests we were submitting. Additionally, we grappled with the task of managing corrupted image files within our dataset, complex issues to troubleshoot. These hurdles, though formidable, became valuable opportunities for growth. Troubleshooting these challenges not only deepened our understanding of Machine Learning, Web development and API integration, but the process of working together to solve these difficulties also served as a source of motivation to overcome these obstacles as a team. It fueled our determination to advance our project to new heights and to develop a tool that could truly make a positive impact in the realms of sustainability and eco-conscious food practices.

Accomplishments that we're proud of

Getting the actual data from the website was a big step for us after multiple website blockings, crashes and unsuccessful attempts. Additionally, developing and training the computer vision model so that it could recognize a random image of a pizza was something that we were very proud of, especially considering it was built in less than 24 hours and we had very little experience working with computer vision before.

What we learned

What an API is, the different basic architectures, particularly REST APIs, the functions used for this architecture How to use an API database; how to actually get the data, find the exact values we were looking for, and then store this data; how to use the request library to accomplish this How to use common gateway interface to search through the Open Food Facts database using the name of the product The basics of machine learning How hard it is to build a deep learning model, especially one that is not implemented through transfer learning. Not only does it take lots of training time, but getting all of the components to mesh is highly difficult

What's next for EcoNibbles

Our next steps is to optimize the machine learning model so it is more accurate when detecting the dish in the picture, and optimizing web scraping so the time it takes to calculate the CO2 emissions is minimized. Furthermore, the website can also be improved by adding more features, and improving the layout. Additional metrics of environmental sustainability , such as carbon footprint, can also be added. The next major steps would be to make an app version which can be used on a mobile phone.