Upload 2 files

app.py

@@ -47,16 +47,24 @@ def get_image_text(image_files):
         try:
             # Open image using PIL
             image = Image.open(image_file)
-            # Convert to numpy array for easyocr
-            image_np = np.array(image)
+            
+            # Preprocess image for better OCR (convert to grayscale and enhance contrast)
+            image = image.convert('L') # Grayscale
+            # Simple contrast enhancement using numpy
+            img_np = np.array(image)
+            # Normalize
+            img_np = (img_np - np.min(img_np)) / (np.max(img_np) - np.min(img_np)) * 255
+            img_np = img_np.astype(np.uint8)
+            
             # Extract text using EasyOCR with parameters optimized for math/handwriting
             results = reader.readtext(
-                image_np,
+                img_np,
                 detail=1,
                 paragraph=False,
                 contrast_ths=0.1,
                 adjust_contrast=0.5,
-                text_threshold=0.5
+                text_threshold=0.4, # Lower threshold to catch faint handwriting
+                low_text=0.3
             )
             # Combine all detected text with confidence filtering
             extracted_text = ' '.join([result[1] for result in results if result[2] > 0.3])
@@ -110,21 +118,42 @@ def render_math_as_image(text):
     Returns tuple: (has_math, processed_text_or_image)
     """
     # Check if text contains LaTeX math patterns
-    math_patterns = [r'\$\$.*?\$\$', r'\$.*?\$', r'\\frac', r'\\sqrt', r'\\sum', r'\\int', r'\\begin\{.*?\}', r'\\\[.*?\\\]']
+    math_patterns = [
+        r'\$\$.*?\$\$', r'\$.*?\$', 
+        r'\\frac', r'\\sqrt', r'\\sum', r'\\int', 
+        r'\\begin\{.*?\}', r'\\\[.*?\\\]',
+        r'\[\s*\\begin'  # Catch [\begin cases
+    ]
     has_complex_math = any(re.search(pattern, text, re.DOTALL) for pattern in math_patterns)
     
-    if has_complex_math and len(text) < 1000:  # Only render as image if not too long
+    if has_complex_math:
         try:
+            # Clean the text for rendering
+            clean_text = text
+            
+            # Normalize delimiters
+            clean_text = clean_text.replace('$$', '$')
+            clean_text = clean_text.replace(r'\[', '$').replace(r'\]', '$')
+            clean_text = clean_text.replace(r'\(', '$').replace(r'\)', '$')
+            
+            # Fix specific issue seen in screenshot: [\begin -> $\begin
+            clean_text = re.sub(r'\[\s*\\begin', r'$\n\\begin', clean_text)
+            clean_text = re.sub(r'\\end\{.*?\}\s*\]', r'\\end{cases}$', clean_text) # Simplified fix for end
+            
+            # Remove spacing arguments like \\[2mm] -> \\
+            clean_text = re.sub(r'\\\\\[.*?\]', r'\\\\', clean_text)
+            
+            # Estimate height based on text length (rough approximation)
+            lines = clean_text.count('\n') + (len(clean_text) // 60)
+            fig_height = max(4, lines * 0.5)
+            
             # Create figure
-            fig, ax = plt.subplots(figsize=(10, 6))
+            fig, ax = plt.subplots(figsize=(10, fig_height))
             ax.axis('off')
             
-            # Clean the text for rendering
-            clean_text = text.replace('$$', '$').replace('\\[', '$').replace('\\]', '$')
-            
             # Render text with math
             ax.text(0.5, 0.5, clean_text, 
-                   fontsize=14, 
+                   fontsize=12, 
                    ha='center', 
                    va='center',
                    wrap=True,
@@ -132,13 +161,14 @@ def render_math_as_image(text):
             
             # Save to bytes
             buf = BytesIO()
-            plt.savefig(buf, format='png', bbox_inches='tight', dpi=150, facecolor='#f5f2e8')
+            plt.savefig(buf, format='png', bbox_inches='tight', dpi=200, facecolor='#f5f2e8')
             buf.seek(0)
             plt.close(fig)
             
             return True, buf
         except Exception as e:
-            # If rendering fails, return text
+            # If rendering fails, return text but try to clean it for markdown
+            print(f"Rendering failed: {e}")
             return False, text
     
     return False, text