app.py

import streamlit as st
from transformers import pipeline, AutoImageProcessor
from PIL import Image, ImageDraw
import torch

st.set_page_config(
    page_title="Fraktur Detektion",
    layout="wide",
    initial_sidebar_state="collapsed"
)

st.markdown("""
<style>
    .stApp {
        background-color: transparent !important;
        padding: 0 !important;
    }
    
    [data-theme="light"] {
        --background-color: #ffffff;
        --text-color: #1f2937;
        --border-color: #e5e7eb;
        --button-color: #2563eb;
        --button-hover: #1d4ed8;
    }
    
    [data-theme="dark"] {
        --background-color: #1f2937;
        --text-color: #f3f4f6;
        --border-color: #4b5563;
        --button-color: #3b82f6;
        --button-hover: #2563eb;
    }
    
    .stButton > button {
        background-color: var(--button-color) !important;
        color: white !important;
        border: none !important;
        padding: 0.5rem 1rem !important;
        border-radius: 0.375rem !important;
        font-weight: 500 !important;
        transition: background-color 0.2s !important;
    }
    
    .stButton > button:hover {
        background-color: var(--button-hover) !important;
    }
    
    .block-container {
        padding: 0.5rem !important;
        max-width: 100% !important;
    }
    
    .stImage > img {
        max-height: 250px !important;
        width: auto !important;
        margin: 0 auto !important;
    }
    
    .result-box {
        padding: 0.375rem;
        border-radius: 0.375rem;
        margin: 0.25rem 0;
        background: var(--background-color);
        border: 1px solid var(--border-color);
        color: var(--text-color);
    }
    
    h2, h3, h4 {
        margin: 0.5rem 0 !important;
        color: var(--text-color) !important;
        font-size: 1rem !important;
    }
    
    #MainMenu, footer, header {
        display: none !important;
    }
    
    .uploadedFile {
        border: 1px dashed var(--border-color);
        border-radius: 0.375rem;
        padding: 0.25rem;
    }
    
    .row-widget.stButton {
        text-align: center;
        margin: 1rem 0;
    }
    
    div[data-testid="stFileUploader"] {
        width: 100%;
    }
</style>
""", unsafe_allow_html=True)

@st.cache_resource
def load_models():
    return {
        "KnochenAuge": pipeline("object-detection", model="D3STRON/bone-fracture-detr"),
        "KnochenWächter": pipeline("image-classification", 
            model="Heem2/bone-fracture-detection-using-xray",
            image_processor=AutoImageProcessor.from_pretrained("Heem2/bone-fracture-detection-using-xray")),
        "RöntgenMeister": pipeline("image-classification", 
            model="nandodeomkar/autotrain-fracture-detection-using-google-vit-base-patch-16-54382127388",
            image_processor=AutoImageProcessor.from_pretrained("nandodeomkar/autotrain-fracture-detection-using-google-vit-base-patch-16-54382127388"))
    }

def draw_boxes(image, predictions):
    draw = ImageDraw.Draw(image)
    for pred in predictions:
        if pred['label'].lower() == 'fracture' and pred['score'] > 0.6:
            box = pred['box']
            label = f"Fraktur ({pred['score']:.2%})"
            color = "#2563eb"
            
            draw.rectangle(
                [(box['xmin'], box['ymin']), (box['xmax'], box['ymax'])],
                outline=color,
                width=2
            )
            
            text_bbox = draw.textbbox((box['xmin'], box['ymin']-15), label)
            draw.rectangle(text_bbox, fill=color)
            draw.text((box['xmin'], box['ymin']-15), label, fill="white")
    return image

def main():
    models = load_models()
    
    st.markdown("### 📤 Röntgenbilder Upload")
    
    # File uploader avec label caché
    uploaded_files = st.file_uploader(
        "Wählen Sie Röntgenbilder aus",
        type=['png', 'jpg', 'jpeg'],
        accept_multiple_files=True,
        label_visibility="collapsed"
    )
    
    if uploaded_files:
        # Bouton d'analyse
        if st.button("Bilder analysieren", key="analyze_button"):
            col1, col2 = st.columns([1, 1])
            
            for idx, uploaded_file in enumerate(uploaded_files):
                image = Image.open(uploaded_file)
                
                # Analyse avec KnochenAuge (localisierung)
                predictions = models["KnochenAuge"](image)
                fractures_found = any(p['label'].lower() == 'fracture' and p['score'] > 0.6 for p in predictions)
                
                # Afficher uniquement si des fractures sont détectées
                if fractures_found:
                    with col1 if idx % 2 == 0 else col2:
                        result_image = image.copy()
                        result_image = draw_boxes(result_image, predictions)
                        st.image(result_image, caption=f"Bild {idx + 1}", use_column_width=True)
                        
                        # Analyse KnochenWächter et RöntgenMeister
                        pred_wachter = models["KnochenWächter"](image)[0]
                        pred_meister = models["RöntgenMeister"](image)[0]
                        
                        if pred_wachter['score'] > 0.6 or pred_meister['score'] > 0.6:
                            st.markdown(f"""
                                <div class='result-box'>
                                    <span style='color: #2563eb'>KnochenWächter:</span> {pred_wachter['score']:.1%}<br>
                                    <span style='color: #2563eb'>RöntgenMeister:</span> {pred_meister['score']:.1%}
                                </div>
                            """, unsafe_allow_html=True)

    # Script pour la synchronisation du thème
    st.markdown("""
        <script>
            // Fonction pour mettre à jour le thème
            function updateTheme(isDark) {
                document.documentElement.setAttribute('data-theme', isDark ? 'dark' : 'light');
                
                // Mise à jour des styles en fonction du thème
                const root = document.documentElement;
                if (isDark) {
                    root.style.setProperty('--background-color', '#1f2937');
                    root.style.setProperty('--text-color', '#f3f4f6');
                    root.style.setProperty('--border-color', '#4b5563');
                } else {
                    root.style.setProperty('--background-color', '#ffffff');
                    root.style.setProperty('--text-color', '#1f2937');
                    root.style.setProperty('--border-color', '#e5e7eb');
                }
            }
            
            // Écouter les messages du parent
            window.addEventListener('message', function(e) {
                if (e.data.type === 'theme-change') {
                    updateTheme(e.data.theme === 'dark');
                }
            });
            
            // Thème initial basé sur les préférences système
            updateTheme(window.matchMedia('(prefers-color-scheme: dark)').matches);
        </script>
    """, unsafe_allow_html=True)

if __name__ == "__main__":
    main()