Color Code Adherence in AI Image Generation: A Case Study on Gemini

This repository contains a two-part case study on achieving precise color adherence with Google's Gemini 2.5 Flash Image (Nano Banana) model. The project includes:

• The Generation Notebook (generation.ipynb): A practical, interactive tool that demonstrates how to use the color Image-reference method.
• The Evaluation Notebook (evaluation.ipynb): A 50-sample quantitative analysis that evaluates the accuracy of the color image-reference method compared to color text-only instructions.

---

📋 Prerequisites

• Python 3.7 or higher
• Jupyter Notebook or JupyterLab
• Google Gemini API key

🚀 How to Use

1. Requirements

You must have Python 3 and Jupyter installed. First, install the dependencies:

pip install -r requirements.txt

2. Enable the ipywidgets Extension for Jupyter

You will also need to enable the ipywidgets extension for Jupyter:

jupyter nbextension enable --py widgetsnbextension

3. Get a Gemini API Key

Get your API key from Google AI Studio.

4. Run the Notebooks

• For the generation notebook (interactive_generator.ipynb):
  Open and run this notebook. The interactive widgets will allow you to generate images, test the "Use color as a reference" feature, and download your results.

• For the evaluation notebook (evaluation_notebook.ipynb):
  Open and run all cells from top to bottom. The 50-sample loop (Cell 8) will take several minutes.

📁 Project Structure

.
├── generation.ipynb              # Interactive image generation tool
├── evaluation.ipynb              # Quantitative evaluation notebook
├── requirements.txt              # Python dependencies
└── README.md                     # This file

🔍 What's Inside

Generation Notebook

• Interactive widgets for real-time image generation
• Color reference image upload functionality
• Side-by-side comparison of text-only vs. image-reference methods
• Download capabilities for generated images

Evaluation Notebook

• Automated testing across 50 sample cases
• Quantitative color accuracy metrics
• Statistical analysis of color adherence
• Visualization of results

📄 License

This project is open source and available under the MIT License.