PhotogChauffeur is a Python-based tool that allows users to search through a collection of images in a specified folder using natural language queries. The program utilizes OpenAI's GPT-4 Vision API to visualize the input images and return relevant images based on the query. Results can be saved to a designated folder for further use. Also has features for image and video description generation.
- "Which images have more than one person in them?"
- "Show me all images with dim lighting."
- "Provide me with all images where people are wearing hats."
- "Which of these images do you think is best?"
- Image search using natural language queries
- Support for processing multiple images concurrently
- Option to save search results to a specified folder
- High-detail image description
- Video description utilizing frame sampling
- Python 3.11
- OpenAI API key for a Plus account
- Clone the repository:
git clone https://github.com/maquenneville/PhotogChauffeur.git - Navigate to the cloned repository:
cd PhotogChauffeur - Install the required dependencies:
pip install -r requirements.txt - Update the config.ini file with your personal OpenAI API key
Run the program using Python:
python vision_search_ui.py
Follow the on-screen prompts to enter the folder paths, queries, image and video files. Type 'exit' to terminate the program.
This tool is not perfect at applying the search query, and is only as powerful as the machine vision capabilities of gpt-4o. It may take multiple attempts with different phrasing, or simply not be able to find images with certain parameters. Experiment to find what works best.
- Image description for single images
- Video description using a sample of frames
- Iterative search, allowing the user to filter photos further after the initial search
Distributed under the MIT License. See LICENSE for more information.