Update app.py

app.py

@@ -6,6 +6,7 @@ import time
 import io
 import subprocess
 import sys
+import cv2
 
 # Install required packages
 def install_packages():
@@ -13,7 +14,8 @@ def install_packages():
         "transformers",
         "accelerate", 
         "timm",
-        "easyocr"
+        "easyocr",
+        "opencv-python"
     ]
     for package in packages:
         try:
@@ -135,17 +137,49 @@ Generate a complete, professional SOAP note:"""
     except Exception as e:
         return f"❌ SOAP generation failed: {str(e)}"
 
+def preprocess_image_for_ocr(image):
+    """Preprocess image for better OCR results using CLAHE"""
+    try:
+        if hasattr(image, 'convert'):
+            image = image.convert('RGB')
+        img_array = np.array(image)
+        
+        # Convert to grayscale
+        if len(img_array.shape) == 3:
+            gray = cv2.cvtColor(img_array, cv2.COLOR_RGB2GRAY)
+        else:
+            gray = img_array
+        
+        # Resize if too small
+        height, width = gray.shape
+        if height < 300 or width < 300:
+            scale = max(300/height, 300/width)
+            new_height = int(height * scale)
+            new_width = int(width * scale)
+            gray = cv2.resize(gray, (new_width, new_height), interpolation=cv2.INTER_CUBIC)
+        
+        # Enhance image with CLAHE
+        gray = cv2.medianBlur(gray, 3)
+        clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
+        gray = clahe.apply(gray)
+        _, gray = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
+        
+        return gray
+    except Exception as e:
+        print(f"⚠️ Image preprocessing failed: {e}")
+        # Fallback to original image if preprocessing fails
+        return np.array(image)
+
 def extract_text_from_image(image):
-    """Extract text using EasyOCR"""
+    """Extract text using EasyOCR with CLAHE preprocessing"""
     if ocr_reader is None:
         return "❌ OCR not available"
     
     try:
-        if hasattr(image, 'convert'):
-            image = image.convert('RGB')
-        img_array = np.array(image)
+        # Apply CLAHE preprocessing for better OCR
+        processed_img = preprocess_image_for_ocr(image)
         
-        results = ocr_reader.readtext(img_array, detail=0, paragraph=True)
+        results = ocr_reader.readtext(processed_img, detail=0, paragraph=True)
         if results:
             return ' '.join(results).strip()
         else:
@@ -197,41 +231,35 @@ EKG: ST elevation in leads II, III, aVF"""
     
     with gr.Blocks(title="Medical OCR SOAP Generator", theme=gr.themes.Soft()) as demo:
         
-        
         gr.HTML("""
-<h1>🏥 Medical OCR SOAP Generator - LIVE DEMO</h1>
-<h2>🎯 For Competition Judges - Quick 2-Minute Demo:</h2>
+        <h1>🏥 Medical OCR SOAP Generator - LIVE DEMO</h1>
+        <h2>🎯 For Competition Judges - Quick 2-Minute Demo:</h2>
 
-<div style="background-color: #e6f3ff; padding: 15px; border-radius: 10px; margin: 10px 0;">
-<h3>📋 TWO DEMO OPTIONS:</h3>
-<p><strong>Option A (OCR Demo):</strong> Download "docs-note-to-upload.jpg" from Files tab above, then upload it below</p>
-<p><strong>Option B (Text Demo):</strong> Click "Try Sample Medical Text" button for instant text-to-SOAP demo</p>
-</div>
+        <div style="background-color: #e6f3ff; padding: 15px; border-radius: 10px; margin: 10px 0;">
+        <h3>📋 SAMPLE IMAGE PROVIDED:</h3>
+        <p><strong>👆 Download "docs-note-to-upload.jpg" from the Files tab above, then upload it below</strong></p>
+        <p><strong>OR</strong> click "Try Sample Medical Text" button for instant text demo</p>
+        </div>
 
-<h3>Demo Steps:</h3>
-<ol>
-<li><strong>Choose your demo:</strong>
-   <ul>
-   <li><strong>OCR Demo:</strong> Upload the sample image → Shows OCR extraction + SOAP generation</li>
-   <li><strong>Text Demo:</strong> Click sample text button → Shows SOAP generation only</li>
-   </ul>
-</li>
-<li><strong>Click "Generate SOAP Note"</strong></li>
-<li><strong>Wait ~2-3 minutes</strong> for AI processing (model loading + generation)</li>
-<li><strong>See professional SOAP note</strong> generated by Gemma 3n</li>
-</ol>
+        <h3>Demo Steps:</h3>
+        <ol>
+        <li><strong>Upload the sample image</strong> (docs-note-to-upload.jpg from Files tab) <strong>OR</strong> click sample text button</li>
+        <li><strong>Click "Generate SOAP Note"</strong></li>
+        <li><strong>Wait ~60-90 seconds</strong> for AI processing (first time only)</li>
+        <li><strong>See professional SOAP note</strong> generated by Gemma 3n</li>
+        </ol>
 
-<h3>✅ What This Demo Shows:</h3>
-<ul>
-<li><strong>Real OCR</strong> extraction from handwritten medical notes (image upload)</li>
-<li><strong>AI-powered medical reasoning</strong> with Gemma 3n (both options)</li>
-<li><strong>Professional SOAP formatting</strong> (Subjective, Objective, Assessment, Plan)</li>
-<li><strong>HIPAA-compliant</strong> local processing</li>
-</ul>
+        <h3>✅ What This Demo Shows:</h3>
+        <ul>
+        <li><strong>Real OCR</strong> extraction from handwritten medical notes</li>
+        <li><strong>AI-powered medical reasoning</strong> with Gemma 3n</li>
+        <li><strong>Professional SOAP formatting</strong> (Subjective, Objective, Assessment, Plan)</li>
+        <li><strong>HIPAA-compliant</strong> local processing</li>
+        </ul>
 
-<p><strong>⚠️ Note:</strong> First generation takes ~2-3 minutes as model loads. Subsequent ones are faster.</p>
-<hr>
-""")
+        <p><strong>⚠️ Note:</strong> First generation takes ~60-90 seconds as model loads. Subsequent ones are faster.</p>
+        <hr>
+        """)
         
         with gr.Row():
             with gr.Column():