Food Nutrition Detector

Overview

Food Nutrition Detector is a small ASP.NET Core (.NET 8) Web API that uses Azure OpenAI to identify food in an uploaded image and estimate basic nutritional values. The API accepts an image and returns a JSON object with the detected food name and nutrient estimates (calories, carbs, protein, fat) along with optional notes.

Key components

• Food_Nutrition_Detector.Services.OpenAiNutritionService -- wraps Azure AI OpenAI client to send an image and prompt to a deployed model and parse the JSON response.
• Food_Nutrition_Detector.Controllers.NutritionController -- exposes a single analyze endpoint that accepts an image upload.
• Food_Nutrition_Detector.Models.NutritionInfo -- the response model returned by the analyze endpoint.

Prerequisites

• .NET 8 SDK
• An Azure OpenAI resource with a deployment that accepts chat + image input
• An API key and endpoint for the Azure OpenAI resource

Configuration

Add your Azure OpenAI settings to appsettings.json (or appsettings.Development.json) under the AzureOpenAI section. Example:

{ "AzureOpenAI": { "Endpoint": "https://YOUR_RESOURCE.openai.azure.com", "ApiKey": "YOUR_API_KEY", "Deployment": "your-deployment-name" } }

How to run

1. Restore packages and run the project using dotnet (or open in Visual Studio).
2. When running in Development the project exposes Swagger UI at /swagger for manual testing.

API: Analyze Image

• Endpoint: POST /api/nutrition/analyze
• Content type: multipart/form-data
• Form field: file (the image to analyze)

Request example (multipart/form-data):

• file: (binary image file)

Response

The API returns a JSON document matching the NutritionInfo model:

{ "foodName": "string", // Detected food name "calories": 0.0, // Estimated calories (kcal) "carbs": 0.0, // Estimated carbs (grams) "protein": 0.0, // Estimated protein (grams) "fat": 0.0, // Estimated fat (grams) "notes": "string" // Additional info or parsing errors }

Notes and error handling

• If the AI response cannot be parsed as JSON, the API will still return a NutritionInfo object with the Notes property populated to indicate the parsing failure and include the raw AI text.
• The accuracy of the estimates depends on the model and the image quality. Treat returned values as rough estimates only.

Extending or customizing

• Update the prompt in OpenAiNutritionService to refine the JSON shape or request units, confidence, or additional nutrients.
• If you want richer responses, consider altering NutritionInfo and the expected JSON format in the service deserialization.