tuandunghcmut/mllm-dla-demo-gradio

MLLM Document Layout Analysis Demo

This Streamlit application demonstrates the capabilities of Multimodal Large Language Models (MLLMs) for Document Layout Analysis (DLA).

Features

• —Upload PDFs or images.
• —Select from multiple SOTA MLLMs (via Hugging Face Inference API).
• —Customizable inference prompt.
• —Visualizes bounding boxes, labels, and entity relations.

Setup

1. 1.Install dependencies:

   pip install -r requirements.txt

1. 1.Install poppler-utils (required by pdf2image):
2. 2.On Linux: sudo apt-get install poppler-utils
3. 3.On Mac: brew install poppler
4. 4.On Windows: Download from http://blog.alivate.com.au/poppler-windows/ and add bin to PATH.

Running the App

streamlit run app.py