PyWebTagger - WIP

This is a standalone dataset creation tool for stable diffusion primarily focused on using tags to describe images.

Getting Started

You can install this program by using Git bash and the following command

git clone https://github.com/SesuMoe/PyWebTagger.git

Once installed run the setup.bat and wait for the installation to finish. Then use the start.bat and go to http://localhost:3000 in your browser.

Note: This is meant for local usage only. Use caution if you attempt to expose the webserver to the public, as you may risk exposing your personal data.

TODO Shell

Usage

Dataset

Provide the path to your dataset, and it will recursively find matching image / caption pairs (e.g. example.txt / example.png). It does not support captions in the filename.

Tags

TODO

Goals

These are some of the long term goals I wish to achieve while working on this project.

• Booru Scrapping - Able to fetch images directly and add them to the dataset.
• Reverse-Image Tag Search - Uses a reverse image service to find and fetch tags from the internet.
• Drag and Drop - Add images to the dataset by dragging and dropping into the UI.
• Interrogator - Ability to inference images to make captioning faster. Done.
• Cropping - Able to crop images whether this is on another page or using the display.
• Auto-Cropping - Set and go cropping with face detection if possible.
• Bulk Tag Editor - Edit sets of tags at a time.
• Gallery View - View multiple images Started.