import streamlit as st

def show():
    st.markdown('<div class="main-header">🏥 Video Surgical Scene Understanding Dashboard</div>', unsafe_allow_html=True)
    st.markdown("---")

    # Welcome and overall description
    st.markdown("## Welcome to the Surgical Scene Analysis Platform")
    st.markdown("""
    This platform demonstrates an end-to-end pipeline for automated understanding of surgical scenes from video data.
    The system leverages advanced computer vision and AI models to analyze surgical workflows, recognize tools, and generate scene-level captions.
    Navigate through the sidebar to test the system, explore datasets, or learn more about the project.
    """)

    st.markdown("---")
    st.markdown("## 🔄 Pipeline Overview")
    st.markdown("""
    The surgical scene understanding pipeline consists of the following main steps:
    1. **Frame Extraction**: Select or upload three consecutive frames from a surgical video.
    2. **Segmentation**: Use the SwinUNETR model to generate a segmentation mask for the scene.
    3. **Captioning**: Input the frames and mask into the MedGemma model to generate a descriptive caption or scene graph.
    4. **Results & Analysis**: Review the generated mask and caption to understand the surgical context.
    """)

    st.markdown("---")
    st.markdown("## 📚 Project Description")
    st.markdown("""
    This project was developed by **Team SATOR** for the ACVSS 2025 Hackathon.  
    Our goal is to provide an accessible, interactive demonstration of state-of-the-art surgical scene understanding using deep learning.
    - **Frontend**: Streamlit Dashboard
    - **Backend**: Python, PyTorch, MONAI, HuggingFace Transformers
    - **Models**: SwinUNETR (segmentation), MedGemma (captioning)
    - **Dataset**: MM-OR (Multimodal Operating Room)
    """)

    st.markdown("---")
    st.info("Use the sidebar to start testing the system or to learn more about the dataset and team.")