Update app.py

app.py

@@ -14,6 +14,10 @@ from scipy.spatial import distance as dist
 # App Configuration & Model Loading
 # ==============================================================================
 
+# For Hugging Face Spaces deployment, you also need these two files:
+# 1. requirements.txt (listing all Python libraries)
+# 2. packages.txt (containing the line "tesseract-ocr")
+
 # Set Streamlit page configuration
 st.set_page_config(
     page_title="Document Scanner & Table Recognizer",
@@ -34,9 +38,8 @@ def load_model():
 processor, model = load_model()
 
 # ==============================================================================
-# Core Image Processing Functions (Slightly adapted for Streamlit)
+# Core Image Processing Functions
 # ==============================================================================
-# (These are the helper functions from the original script)
 
 def order_points(pts):
     xSorted = pts[np.argsort(pts[:, 0]), :]
@@ -77,8 +80,15 @@ def correct_orientation(image):
     try:
         osd = pytesseract.image_to_osd(image, output_type=pytesseract.Output.DICT)
         rotation = osd['rotate']
-        if rotation > 0:
-            rotated_image = cv2.rotate(image, eval(f'cv2.ROTATE_{rotation}_CLOCKWISE'))
+        if rotation in [90, 180, 270]:
+            # The rotation values from Tesseract are counter-clockwise.
+            # OpenCV's rotation constants are clockwise. We need to map them correctly.
+            if rotation == 90:
+                rotated_image = cv2.rotate(image, cv2.ROTATE_90_COUNTERCLOCKWISE)
+            elif rotation == 180:
+                rotated_image = cv2.rotate(image, cv2.ROTATE_180)
+            else: # 270
+                rotated_image = cv2.rotate(image, cv2.ROTATE_90_CLOCKWISE)
             return rotated_image
     except Exception as e:
         st.warning(f"OSD check failed: {e}. Returning original image.")
@@ -88,7 +98,7 @@ def extract_and_draw_table_structure(image_bgr):
     """Takes a BGR image, finds table structure, and returns an image with boxes."""
     image_pil = Image.fromarray(cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB))
     inputs = processor(images=image_pil, return_tensors="pt")
-    
+
     with torch.no_grad():
         outputs = model(**inputs)
 
@@ -96,7 +106,6 @@ def extract_and_draw_table_structure(image_bgr):
     target_sizes = torch.tensor([image_pil.size[::-1]])
     results = processor.post_process_object_detection(outputs, threshold=0.7, target_sizes=target_sizes)[0]
 
-    # Use Matplotlib to draw on the image
     fig, ax = plt.subplots(1, figsize=(width / 100, height / 100), dpi=100)
     ax.imshow(image_pil)
     ax.axis('off')
@@ -110,7 +119,8 @@ def extract_and_draw_table_structure(image_bgr):
             ax.add_patch(rect)
 
     fig.canvas.draw()
-    img_with_boxes = np.frombuffer(fig.canvas.tostring_rgb(), dtype=np.uint8)
+    # FIX: Replaced deprecated 'tostring_rgb' with 'tobytes_rgb'
+    img_with_boxes = np.frombuffer(fig.canvas.tobytes_rgb(), dtype=np.uint8)
     img_with_boxes = img_with_boxes.reshape(fig.canvas.get_width_height()[::-1] + (3,))
     plt.close(fig)
     return img_with_boxes
@@ -125,32 +135,29 @@ st.write("Upload a document photo. The app will automatically straighten it, fix
 uploaded_file = st.file_uploader("Choose a document image...", type=["jpg", "jpeg", "png"])
 
 if uploaded_file is not None:
-    # Convert the uploaded file to an OpenCV image
     file_bytes = np.asarray(bytearray(uploaded_file.read()), dtype=np.uint8)
-    input_image = cv2.imdecode(file_bytes, 1) # 1 for color
+    input_image = cv2.imdecode(file_bytes, 1)
     input_image_rgb = cv2.cvtColor(input_image, cv2.COLOR_BGR2RGB)
 
     st.subheader("1. Original Image")
-    st.image(input_image_rgb, caption="Your Uploaded Image", use_column_width=True)
+    # FIX: Replaced deprecated 'use_column_width' with 'use_container_width'
+    st.image(input_image_rgb, caption="Your Uploaded Image", use_container_width=True)
 
     with st.spinner("Processing your document... This may take a moment."):
-        # Step 1: Straighten and orient the document
         straightened_image = find_and_straighten_document(input_image)
-        image_to_process = straightened_image if straightened_image is not None else input_image
+        image_to_process = straightened_image if straightened_image is not None and straightened_image.size > 0 else input_image
         final_image = correct_orientation(image_to_process)
         final_image_rgb = cv2.cvtColor(final_image, cv2.COLOR_BGR2RGB)
 
-        # Step 2: Extract table structure
         image_with_structure = extract_and_draw_table_structure(final_image)
 
     st.subheader("2. Corrected Document & Detected Structure")
     col1, col2 = st.columns(2)
 
     with col1:
-        st.image(final_image_rgb, caption="Auto-Corrected & Oriented", use_column_width=True)
-        
-        # Create a download button for the clean image
-        # Convert to bytes for the download button
+        # FIX: Replaced deprecated 'use_column_width' with 'use_container_width'
+        st.image(final_image_rgb, caption="Auto-Corrected & Oriented", use_container_width=True)
+
         _, buf = cv2.imencode(".jpg", final_image)
         st.download_button(
             label="Download Clean Image",
@@ -160,4 +167,5 @@ if uploaded_file is not None:
         )
 
     with col2:
-        st.image(image_with_structure, caption="Detected Table Structure (Rows: Green, Columns: Red)", use_column_width=True)
+        # FIX: Replaced deprecated 'use_column_width' with 'use_container_width'
+        st.image(image_with_structure, caption="Detected Table Structure (Rows: Green, Columns: Red)", use_container_width=True)