CARBS Processing Backend

Overview

The CARBS Processing Backend is the server-side component of the CARBS (Carbohydrate Analysis and Recognition By Scan) system. It processes images captured by the Android app to perform food detection, volume estimation using fiducial markers, and carbohydrate calculation, returning results to the client app.

View the CARBS Android Capture Tool app here!

Features

• Image upload and food detection via a YOLOv2 model trained on UEC-Food100.
• Volume estimation using ArUco fiducial markers.
• Conversion of food volume to weight and carbohydrate estimation.
• Results bundling and response delivery to client app.
• Deployable locally or via cloud services like Microsoft Azure.

Technologies

• Python 3.x
• Flask
• OpenCV
• YOLOv2
• UEC-Food100 dataset

API Endpoints

Method	Endpoint	Description
GET	/test	Test server connectivity
POST	/upload-data	Upload an image for processing
GET	/get-food-classes	Retrieve list of supported food classes

Deployment

• Backend has been successfully deployed to Azure for remote access.
• HTTPS configuration and resource scaling may be needed for production use.