CalorAI

Logo

Inspiration

Struggling to track calories without access to precise scales was a common issue among myself and those in my circle. This challenge inspired the creation of a solution that could overcome the limitations posed by the absence of accurate measuring tools.

What it Does

CalorAI is an innovative app designed to measure food volume, calculate mass by multiplying it with density, and then determine the calorie content of the food. Leveraging advanced technologies like PyTorch, TensorFlow, Swift, OpenCV, LiDAR distortion correction, LiDAR contour search, FastAPI, and the Render deployment platform, CalorAI simplifies calorie tracking without the need for precise scales.

How We Built It

Our development process involved using PyTorch for food classification and TensorFlow for semantic segmentation, Swift for app development, and OpenCV for image processing of the Lidar output. We used modified and implemented LiDAR distortion correction technology and accurate search for contours. The integration of FastAPI facilitated seamless communication between the app and the ML models. However, one of our main challenges revolved around uploading the ML API to a hosting server for app use. We used Render free hosting server. Also we used Wolfram Alpha's API for calculating calories in food.

Challenges We Ran Into

Throughout the project, we faced several hurdles, notably involving the upload of the ML API to the hosting server for app integration. Developing an accurate algorithm to calculate volume using LiDAR and AI presented its complexities, especially in cleaning and segmenting complex data. Additionally, integrating the API with the Swift mobile app demanded significant effort.

Accomplishments That We're Proud Of

Our proudest achievement lies in achieving a high level of accuracy in calculating calories with a single picture and LiDAR data. We enhanced the app's effectiveness in empowering users to make informed dietary choices.

What We Learned

Our journey with CalorAI was a learning experience. We gained insights into using LiDAR technology within a Swift mobile app, mastering the hosting of large machine learning models, and creating custom TensorFlow datasets.

What's Next for CalorAI

Moving forward, our focus lies in fine-tuning both ML models and the volume calculation algorithm. This step aims to further improve accuracy and efficiency, ensuring CalorAI remains at the forefront of simplified and accurate calorie tracking.