Adjusted LLava captioning webui

Thanks to ShadoW-Shinigami for writing mass image captioning based on Llava architecture

Also thanks to gessyo, the adjustment inspired from gessyoo/minigpt4-batch-tweaked, but it's for Llava architecure

The output should be removing unnecessary picture like, "This images shows", "The picture features", and so on. I still need some adjustment to remove some "unnecessary" word that maybe generated by Llava

---

Original Readme

Llava-WEBUI-Caption: Enhance Your Image Captioning Experience

Llava-WEBUI-Caption is a user-friendly Gradio-based interface designed to streamline the process of captioning image datasets using the LLaVA Model. Whether you're dealing with PNG, JPG, JPEG, or WEBP formats, this tool simplifies the task of generating and managing image captions.

🌟 Features

• Image Directory Captioning: Automatically create text captions for images in various formats, saving them alongside the original files.
• Download & Caption: Just add image URLs to an "images.json" file in your source folder and let the script handle the downloading and captioning.
• Caption Management: Choose to skip, replace, or append to existing captions, giving you full control over your captioning process.

⚙️ Installation

1. Clone the LLAVA repository.
2. Follow the instructions provided in the LLaVA repo to set up.
3. Add the Llava-WEBUI-Caption script to the downloaded folder and run it to start captioning your images.