Imagen Product Recontextualization at Scale

This repository contains Jupyter notebooks and tools to perform large-scale product image recontextualization using Google’s Gemini and Imagen Product Recontext models. It includes both the generation pipeline and an evaluation framework.

Authors: Layolin Jesudhass & Isidro De Loera

Contents

Notebooks

• imagen_product_recontext_at_scale.ipynb
  Scales up product image recontextualization using Imagen. Handles:

  • Batch generation of recontextualized product images
  • Prompt engineering for diverse product contexts
  • Sequential and Parallel execution options.

• evaluation_imagen_product_recontext_at_scale.ipynb
  Evaluates the generated images on various axes, such as:

  • Product Fidelity
  • Scene Realism
  • Aesthetic Quality
  • Brand Integrity
  • Policy Compliance
  • Imaging Quality
  • Sequential and Parallel execution options.

Getting Started

Requirements

• Python 3.8+
• Jupyter or VSCode
• Google Cloud Vertex AI and access to Imagen Product Recontext API
• Required Python libraries are listed in requirements.txt.

Installation

1. Clone the repository:

   git clone https://github.com/GoogleCloudPlatform/vertex-ai-creative-studio.git

2. Navigate to the experiment directory:

   cd vertex-ai-creative-studio/experiments/Imagen_Product_Recontext

3. Install the required dependencies:

   pip install -r requirements.txt

Usage

1. Generate Images: Open and run the imagen_product_recontext_at_scale.ipynb notebook to generate recontextualized product images.
2. Evaluate Images: Open and run the evaluation_imagen_product_recontext_at_scale.ipynb notebook to evaluate the generated images.