Academy of Interdisciplinary Studies Division of Integrative Systems and Design 250 Deep Learning for Food Nutrition Supervisor: Mitch LI / ISD Student: CHENG Ziheng / ISD Course: UROP 1100, Fall UROP 2100, Summer This report details the development of an integrated system for automated estimation of food volume and nutritional information using multi-modal deep learning and computer vision techniques. The system processes a single RGB image containing food and a reference coin, utilizing state-of-the-art food semantic segmentation, monocular depth estimation, and robust scale calibration. By fusing these components, the system accurately segments food items, infers real-world dimensions, and computes food volumes, which are then linked to nutrition databases for comprehensive dietary analysis. The project introduces several technical innovations, including adaptive region merging for segmentation masks, self-calibrated depth estimation, and robust ellipse-based scale detection. Experimental evaluation demonstrates the system’s reliability and practical value across varied, real-world scenarios, offering a scalable solution for dietary monitoring and nutritional assessment with minimal manual intervention. Deep Learning for Food Nutrition Supervisor: Mitch LI / ISD Student: YAO Ruixin / IS Course: UROP 1100, Fall UROP 2100, Spring This thesis explores the gaming design and construction of computer gaming in the Unity project, with an intended usage of the newly developed EIT project from the fellow. It’s a 2D-based parkour game with different levels of difficulty that aim to adapt to the strength needed for the recovery process. Furthermore, different scenes and extra labels of enemies with goals are introduced to enhance the playability of the game. Generative AI in 3D Food Printing Supervisor: Mitch LI / ISD Student: SHUKLA Saanvi / COMP Course: UROP 1100, Fall Under the supervision of Assistant Professor Mitch Li, I developed an innovative iOS mobile application for UROP 1100, integrating generative AI to analyze food items via camera input and generate detailed nutritional intake summaries with 90% accuracy in volume estimation. Collaborating with teammate Charles and mentor Siyu Chen, I tackled challenges like debugging over 30 code iterations to resolve “build failed” errors, creating an intuitive UI/UX that improved user interaction by 25% in testing. This project expanded my understanding of AI’s role in nutrition, enhanced my Swift programming skills through 100+ hours of hands-on work, and fostered a resilient, growth-oriented mindset by viewing setbacks as learning opportunities.
RkJQdWJsaXNoZXIy NDk5Njg=