Photo Fitness

Inspiration

Both my partner and I are very health conscious with my partner having trouble losing weight due to inconvenience of tracking calories and me having trouble gaining weight with the inaccuracies of calorie trackers. The inspiration for Photo Fitness came from the need to simplify food tracking for people like us who are looking to improve themselves by losing or gaining weight. Regular food logging apps require manual input, which can be tedious, time consuming and inaccurate. We wanted to create an app that made tracking macros as easy as snapping a photo. This project is a step toward leveraging technology to promote healthier eating habits.

What it does

Photo Fitness allows users to take a picture of their meal, automatically identifies the food using image recognition, and provides detailed macro information such as calories, protein, carbs, and fats. Users can track their meals and macros without the hassle of manual logging.

How we built it

1. Image Recognition: We used a pre-trained Google cloud image recognition model to recognize food items from pictures.
2. Nutritional Data: By integrating with the Nutritionix API, we fetch macro details for the identified food items.
3. User Interface: The frontend was developed using React Native to ensure a seamless mobile experience.
4. Data Storage: Firebase handles user authentication and stores the history of logged meals in the cloud.

Challenges we ran into

• Accuracy of Food Recognition: Differentiating between visually similar food items required fine-tuning the image recognition model.
• API Limitations: The Nutritionix API has rate limits, and obtaining reliable data for less common foods was challenging.
• Ambiguous Foods: Identifying mixed or complex dishes from a single image was tricky, and we had to account for potential inaccuracies in food identification.

Accomplishments that we're proud of

• Successfully implementing an end-to-end system that combines machine learning and nutrition data to solve a real-world problem.
• Building an intuitive mobile interface that makes food tracking easy and accessible.
• Overcoming challenges related to image recognition accuracy and API constraints to deliver reliable macro information.

What we learned

• Machine Learning: We learned about how machine learning interacts with media to produce the desired results.
• API Integration: We handled rate limits and error responses from third-party APIs and connected APIs together to work harmoniously to create a functioning software.
• User Experience: Ensuring the app was easy to use and visually appealing was always a priority.

What's next for Photo Fitness

• Expand Food Database: Improve recognition accuracy by training on larger and more diverse datasets.
• Handle Complex Meals: Introduce manual adjustments for users to select food items in mixed or ambiguous meals.
• Macro Goals: Add features allowing users to set daily macro goals and track their progress over time.
• Cross-Platform Support: Expand the web app to include a mobile app version for broader and easier accessibility.

Built With

• css
• next.js
• react
• typescript