Image-Text-Image Conversion Tool

This tool provides functionality to process images and texts in a bidirectional manner:

• Generate text descriptions from images (image → text)
• Generate images from text descriptions (text → image)

The tool maintains directory structures throughout conversions, making it easy to process batches of images or texts while preserving their organization.

Table of Contents

• Project Structure
• Configuration
• Environment Setup
• API Key Setup
• Usage
• Author

Project Structure

.
├── data/
│   ├── real/      # Source images directory
│   ├── text/      # Generated text descriptions directory
│   └── output/    # Generated images directory
├── utils/
│   ├── download_image.py
│   ├── text_to_image.py
|   └── ...
├── main.py        # Main execution script
├── keys.json      # API keys configuration
└── README.md

Configuration

The application uses a configuration dictionary in main.py with the following parameters:

config = {
    "override_text_prompt": False,   # Whether to override existing text files
    "override_output_image": True,   # Whether to override existing image files
    "real_image_path": "./data/real", # Path to the source images
    "text_image_path": "./data/text", # Path to the generated text descriptions
    "output_path": "./data/output",   # Path to the generated images
    "text_prompt": "What is the main content of the image? Please generate a detailed prompt to create an image, following the format of artistic style + subject description, for example: ..."
}

You can modify these settings according to your needs:

• Set override_text_prompt to True to regenerate existing text descriptions
• Set override_output_image to True to regenerate existing images
• Customize the text_prompt to get different types of image descriptions

Environment Setup

Install required packages using pip:

# Install the volcengine SDK for ARK runtime
pip install -U volcengine-python-sdk[ark]

# Install the volcengine Python SDK
pip install --user volcengine

API Key Setup

You need three API keys for this application, corresponding to Volcengine, ARK, and Aliyun OSS. Create a keys.json file with the following structure:

{
    "oss": {
        "access_key_id": "aliyun_access_key_id",
        "access_key_secret": "aliyun_access_key_secret",
        "bucket_name": "aliyun_bucket_name",
        "endpoint": "aliyun_endpoint"  // Example: oss-cn-beijing.aliyuncs.com
    },
    "ark": {
        "api_key": "volc_ark_api_key"
    },
    "volc": {
        "ak": "volc_ak",
        "sk": "volc_sk"
    }
}

How to obtain the API keys:

ARK API Key

Get your API key for the doubao-1-5-vision-pro-32k model from Volcengine ARK Console.

Volcengine API Key

Get your API key from the Volcengine IAM Console.

Aliyun OSS API Key

Get your API key from the Aliyun OSS Console.

Usage

Running the Application

1. Ensure your images are placed in the ./data/real directory with any folder structure you want to maintain
2. Run the main script:

   python main.py
   

3. You'll be prompted to choose an action:
   • Option 1: Generate text descriptions from images
   • Option 2: Generate images from text descriptions

Process Flow

Image to Text Conversion:

• Images from ./data/real will be processed
• Text descriptions will be saved to ./data/text with the same directory structure
• Files will be skipped if they exist and override_text_prompt is False

Text to Image Conversion:

• Text files from ./data/text will be processed
• Generated images will be saved to ./data/output with the same directory structure
• Files will be skipped if they exist and override_output_image is False

Author

[email protected]