Create app.py

app.py

@@ -0,0 +1,126 @@
+import gradio as gr
+from transformers import pipeline
+from PIL import Image, ImageDraw
+import torch
+
+def load_models():
+    return {
+        "KnochenAuge": pipeline("object-detection", model="D3STRON/bone-fracture-detr"),
+        "KnochenWächter": pipeline("image-classification", model="Heem2/bone-fracture-detection-using-xray"),
+        "RöntgenMeister": pipeline("image-classification", 
+            model="nandodeomkar/autotrain-fracture-detection-using-google-vit-base-patch-16-54382127388")
+    }
+
+def draw_boxes(image, predictions, conf_threshold=0.6):
+    draw = ImageDraw.Draw(image)
+    fractures_found = False
+    
+    for pred in predictions:
+        if pred['label'].lower() == 'fracture' and pred['score'] >= conf_threshold:
+            fractures_found = True
+            box = pred['box']
+            label = f"Fraktur ({pred['score']:.1%})"
+            color = "#2563eb" if pred['score'] > 0.7 else "#eab308"
+            
+            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 if fractures_found else None
+
+def analyze_images(images, conf_threshold=0.6):
+    models = load_models()
+    results = []
+    
+    for img in images:
+        pil_img = Image.fromarray(img)
+        
+        # KnochenAuge Analysis
+        predictions = models["KnochenAuge"](pil_img)
+        fractures_found = any(p['label'].lower() == 'fracture' and p['score'] >= conf_threshold 
+                            for p in predictions)
+        
+        if fractures_found:
+            # Draw boxes on image
+            result_image = draw_boxes(pil_img.copy(), predictions, conf_threshold)
+            
+            # Additional analyses
+            wachter_pred = models["KnochenWächter"](pil_img)[0]
+            meister_pred = models["RöntgenMeister"](pil_img)[0]
+            
+            if result_image:
+                results.append({
+                    "image": result_image,
+                    "knochen_wachter": f"KnochenWächter: {wachter_pred['score']:.1%}",
+                    "rontgen_meister": f"RöntgenMeister: {meister_pred['score']:.1%}"
+                })
+    
+    # Format results for display
+    if not results:
+        return None, "Keine Frakturen gefunden."
+    
+    output_images = [r["image"] for r in results]
+    analysis_text = "\n\n".join([
+        f"Bild {i+1}:\n{r['knochen_wachter']}\n{r['rontgen_meister']}"
+        for i, r in enumerate(results)
+    ])
+    
+    return output_images, analysis_text
+
+# Interface configuration
+css = """
+.gradio-container {
+    background-color: transparent !important;
+}
+.dark {
+    background-color: #1f2937;
+    color: #f3f4f6;
+}
+.light {
+    background-color: #ffffff;
+    color: #1f2937;
+}
+"""
+
+with gr.Blocks(css=css) as demo:
+    with gr.Row():
+        with gr.Column(scale=1):
+            file_upload = gr.File(
+                label="Röntgenbilder hochladen",
+                file_types=["image"],
+                file_count="multiple"
+            )
+            conf_slider = gr.Slider(
+                minimum=0.0,
+                maximum=1.0,
+                value=0.6,
+                step=0.05,
+                label="Konfidenzschwelle"
+            )
+            analyze_btn = gr.Button("Bilder analysieren", variant="primary")
+        
+        with gr.Column(scale=2):
+            gallery = gr.Gallery(label="Ergebnisse").style(grid=2)
+            analysis_output = gr.Textbox(label="KI-Analyse", lines=4)
+    
+    analyze_btn.click(
+        fn=analyze_images,
+        inputs=[file_upload, conf_slider],
+        outputs=[gallery, analysis_output]
+    )
+
+# Launch configuration
+demo.launch(
+    show_api=False,
+    share=False,
+    server_name="0.0.0.0",
+    server_port=7860,
+    show_error=True,
+    enable_queue=True
+)