[Demo] Gemma CalorAI: Hugging Face Spaces Demo of Gemma 3 Image-Text-to-Text

Hello Gemma team and community members,

Here, I am sharing my latest Hugging Face spaces demo, [Gemma CalorAI](https://huggingface.co/spaces/halfacupoftea/Gemma_Calorai), an AI Calorie tracker app powered by Gemma 3 Image-text-to-text.

## Overview
Gemma CalorAI is a Hugging Face Spaces app that analyzes food images and provides a detailed nutritional breakdown – all powered by the Gemma 3 model. This app is built for demonstrating a specific speciality of Gemma 3 Image-text-to-text model for a niche industry (Health and Nutrition) inspired by [Google DeepMind's project proposals](https://gist.github.com/dynamicwebpaige/92f7739ad69d2863ac7e2032fe52fbad#11-gemma-model-projects-).

## Key Features
Upload a food image, and this app will return:

- The name of the dish
- Estimated calories
- Breakdown of macronutrients: protein, carbs, fat, fiber, sugar
- Itemized list of each food component with its portion and calorie estimate

## Implementation Details
- [Google Gemma 3-4B-IT](https://huggingface.co/google/gemma-3-4b-it) – image-to-text generation
- Transformers pipeline from Hugging Face
- Gradio interface with PWA (Progressive Web Apps) support
- Hosted on Hugging Face Spaces with GPU acceleration

## How to Use?
- Upload an image of your food item.
- Click **"Scan"**.
- See the detailed nutrition breakdown on the right.
- Want to reset? Click **"Clear"** to start fresh.

## PWA Support – Use it Like an App
This space supports PWA (Progressive Web App) features:

1. Open the space in a browser on your phone
2. Tap "Add to Home Screen" (you can find it on three dots)
3. Use it like a native app - no need to visit the browser every time you want to scan a food snap!

## Links
- [Hugging Face Spaces Demo](https://huggingface.co/spaces/halfacupoftea/Gemma_Calorai)
- [GitHub Repo](https://github.com/aditi-dsi/Gemma-CalorAI)