Overview
A multi-stage vision pipeline: YOLO26n-Seg for detection and segmentation, a ViT-LSTM model for confidence-aware recognition refinement, and a ViT-Small model for meal-level nutrition estimation. It reaches an F1-score of 0.9007 on multi-food recognition refinement and a test MAE of 18.05 for meal-level calorie and macronutrient regression. Delivered as a full-stack mobile platform (Expo React Native frontend, FastAPI backend) with customizable dietary profiles for allergy and condition management.
Methods
- Detection and segmentation
- Recognition refinement
- Nutrition regression