Nutrireader

Main menu
Explanation screens

Inspiration

Nutrireader's main source of inspiration is Nutri-Score, a front-of-pack nutrition label used in Europe. Front-of-pack, or FOP, labels provide simplified information about the nutritional value of a food or drink. By delivering this information to consumers, people become more conscious of the nutritional value of their diet, promoting healthier eating habits. This in turn encourages producers to develop healthier foods and drinks. Evidence has shown that FOP labels have had a positive impact on consumer buying behavior with respect to nutrition. And most importantly, the algorithm worked; healthier foods were consistently given higher grades, demonstrating that the system was a truly accurate indicator of nutritional value.

What it does

Based on a scanned-in (with an AI OCR API) or manually-entered nutrition label, Nutrireader assigns food items a score from 0 to 100 using a research-backed algorithm in Europe. Alongside this score, Nutrireader gives an easy-to-understand GPT-3 AI generated explanation of each category of the food or drink’s label and healthier alternatives. Nutrireader fits perfectly into the hackathon’s overall theme of health by tackling such a prevalent issue today.

How we built it

To make the project, we coded in Python with VS Code We used the Tkinter library to make the GUI We used OpenCV to work with webcams We trained an AI OCR (optical character recognition) model with real nutrition label images to recognize text nearly instantly with Mindee We used the OpenAI API to give detailed, accurate explanations with the GPT-3 language model

Challenges we ran into

Creating the GUI, making all fields align in a modern interface that looks and performs well Training the OCR and getting it to work consistently, as nutrition labels are complex tables of information Finding the NLP model, as many models required payment or needed large downloads

Accomplishments that we're proud of

Nutrireader has significant application in the real world. First, its accurate assessment of an item’s nutritional value can lead people to form healthier eating habits. In addition, the explanations provided help users understand why certain foods are healthy or unhealthy. It’s important to understand what you’re consuming beyond only a number after all. Lastly, the suggested alternatives can further promote healthy eating by providing more nutritional options for food and drink. Anyone who pursues healthy and nutritional habits can reap these benefits. After all, nutrition is one of the most important factors in living a healthy life. While Nutrireader‘s inspiration was drawn from Nutri-Score, it addresses many flaws with the original system. One of these flaws is the lack of detail for a given score. For example, a food or drink may have a high sodium content but still be scored highly by Nutri-Score. However, Nutrireader‘s elaboration on the ingredient composition of a food or drink explains to the user why an item receives the grade that it does. As a result, users can make more informed decisions about their eating habits. In addition, Nutri-Score is only common in Europe while Nutrireader can be used anywhere and by anyone. The 0-100 scale can compare different items more easily than Nutri-Score’s A-E. Lastly, products that use the Nutri-Score label only do so by the decision of their producer, while Nutrireader users can generate a grade for any food or drink at any time.

What's next for Nutrireader

Nutrireader can only run on computers, so porting to phones can make it more convenient to get a score. Creating a database with many known foods could also be helpful, and drawing direct comparisons to foods in the same specific category would give additional insight (e.g., seltzer water instead of a soda, or low-sodium soy sauce instead of normal sodium).