How kushal-at/nutriscanai26 works
NutraVision analyzes paired meal photos: the web client submits them to the nutrition API, which orchestrates vision-based component and portion estimation, then predicts and aggregates nutrients. The vision service combines segmentation, plate calibration, local multimodal interpretation, and density-based weight estimation; the nutrition service resolves food features and runs CatBoost models. The frontend also documents diary and dashboard capabilities, while sampled sync code shows optional Supabase authentication and cloud synchronization. Some frontend and service wiring is documented rather than fully sampled, so those links cite the README.
Main entry points: vision/app.py Vision API · ML/app.py Nutrition API · script.js Client controller · predictor.py Nutrition prediction
Components 19
Web client
- Meal UI
FrontEnd/index.html - Client controller
FrontEnd/script.js - Food diary
FrontEnd/script.js
Vision service
- Vision API
vision/app.py - Food segmentation
vision/segmentation.py - Segmentation model
vision/models/FoodSeg103/best_model.pth - Meal interpretation
vision/vlm.py - Portion estimation
vision/estimation.py - Food density data
vision/food_density.json
Nutrition service
- Nutrition API
ML/app.py - API schemas
ML/schemas.py - Nutrition prediction
ML/predictor.py - Food preprocessing
ML/preprocessor.py - CatBoost models
ML/energy_kcal.cbm
Account sync
- Sync engine
FrontEnd/sync.js - Supabase config
FrontEnd/supabase-config.js
Other components
- Meal user
- Supabase
- Ollama Qwen
Connections 18
- Meal user → Meal UI: submits photos
README.md - Meal UI → Client controller: runs interface
README.md - Client controller → Nutrition API: requests analysis
README.md - Nutrition API → Vision API: forwards photos
ML/app.py - Vision API → Food segmentation: segments images
vision/app.py - Food segmentation → Segmentation model: uses weights
- Vision API → Meal interpretation: interprets meal
vision/app.py - Meal interpretation → Ollama Qwen: queries model
vision/vlm.py - Vision API → Portion estimation: estimates portions
vision/app.py - Portion estimation → Food density data: reads density
- Vision API → Nutrition API: returns estimates
README.md - Nutrition API → Nutrition prediction: predicts nutrients
ML/app.py - Nutrition API → API schemas: validates requests
ML/app.py - Nutrition prediction → Food preprocessing: builds features
ML/predictor.py - Nutrition prediction → CatBoost models: loads regressors
ML/predictor.py - Sync engine → Supabase config: reads credentials
FrontEnd/sync.js - Sync engine → Supabase: syncs account data
FrontEnd/sync.js - Client controller → Food diary: manages diary
FrontEnd/script.js