Dynamic Image Advertising Template Generator

This project implements a Python application that leverages Stable Diffusion for image generation, FastAPI for API development, HTML for user interaction, and Docker for deployment on Hugging Face Spaces.

Table of contents

• General Info
• Usage
• Project Structure
• Furher Improvements
• Author

General Info

The project's main objective is to generate new images resembling a provided stock photo using Stable Diffusion's Img2Img pipeline. FastAPI is utilized to create an API endpoint, enabling users to input image, text prompt, and color(hex code), and receiving a generated image in return. The HTML interface facilitates user interaction, displaying the resulting image. Docker is employed for containerization and deployment on Hugging Face Spaces.

Task

• Utilized Stable Diffusion's Img2Img pipeline to generate images resembling a provided stock photo.
• Use the generated image along with other specified inputs to craft a basic dynamic advertising template.
• Created an API endpoint using FastAPI to receive inputs and return generated images.
• Implemented an HTML interface for user interaction and image display.
• Dockerized the application for easy deployment on Hugging Face Spaces.

Important Note

Please note that due to the computational complexity of the image rendering process, it may take approximately 5-10 minutes to generate a single ad template.

Usage

visit the following link: Ad Fast Api Space

Usage Example (Click to enlarge)	Result Example (Click to enlarge)

Project Structure

├── templates
│   ├── ad_display.html
│   ├── get_input.html
├── main.py
├── ad_creator.py
├── README.md
└── requirements.txt

Further Improvements

1. GPU Utilization: Enhance image generation speed by integrating GPU acceleration, leveraging parallel processing power to expedite the creation of images.
2. Stable Diffusion Enhancement: Explore hyperparameter tuning, model selection, ensemble methods, transfer learning, and community research to optimize the Stable Diffusion model for improved image generation results.
3. Image Format Flexibility: Implement the capability to handle multiple image formats such as JPG, BMP, and others, expanding compatibility for image inputs and outputs.
4. Enhanced HTML Hex Code Support: Address limitations with the webcolors library by extending the range of HTML hex code support. This improvement will broaden the application's ability to accurately interpret and utilize a wider spectrum of HTML hex color codes for user inputs.

Author

• Suleyman Erim