Add application file

app.py

@@ -0,0 +1,72 @@
+import cv2
+import os
+import pytesseract
+import gradio as gr
+from gradio import Interface, Image, Text
+import numpy as np
+from PIL import Image
+from PIL import UnidentifiedImageError
+
+def process_image(input_image):
+    try:
+        # Convert the input image to a NumPy array if it's a PIL Image
+        if isinstance(input_image, Image.Image):
+            img = np.array(input_image)
+        else:
+            # If it's a file path or file-like object, read it directly with OpenCV
+            img = cv2.imread(input_image)
+
+        # Check that the image is in the expected format
+        if img is None or img.dtype != np.uint8:
+            raise Exception("Could not read the image. Please check the image format.")
+
+        img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+
+        # img = cv2.imdecode(np.fromstring(input_image.read(), np.uint8), cv2.IMREAD_COLOR)
+        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+        ret, thresh1 = cv2.threshold(gray, 0, 255, cv2.THRESH_OTSU | cv2.THRESH_BINARY_INV)
+        rect_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (18, 18))
+        dilation = cv2.dilate(thresh1, rect_kernel, iterations=1)
+
+        # Find text lines using connected component analysis
+        text_lines = []
+        contours, hierarchy = cv2.findContours(dilation, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+        for cnt in contours:
+            x, y, w, h = cv2.boundingRect(cnt)
+            text_lines.append((y, y + h, x, x + w))
+
+        # Sort text lines by their y-coordinates
+        text_lines.sort(key=lambda line: line[0])
+
+        # Extract text from each line using Tesseract
+        recognized_text = []
+        for y_min, y_max, x_min, x_max in text_lines:
+            cropped_img = img[y_min:y_max, x_min:x_max]
+            custom_config = r'-l eng+khm --oem 3 --psm 6'
+            extracted_text = pytesseract.image_to_string(cropped_img, config=custom_config)
+            recognized_text.append(extracted_text.strip())
+
+        # Combine recognized text into a single string
+        full_text = "\n".join(recognized_text)
+
+        # Draw bounding boxes on the image
+        result_rgb = img.copy()
+        for y_min, y_max, x_min, x_max in text_lines:
+            cv2.rectangle(result_rgb, (x_min, y_min), (x_max, y_max), (0, 255, 0), 2)
+
+        return full_text, result_rgb
+    except Exception as e:
+        return "Could not process the image. Error: " + str(e), None
+
+iface = gr.Interface(
+    process_image,
+    inputs=[gr.Image(type="pil", label="Processed Image")],
+    outputs=[
+        gr.Text(label="Detected Labels"),
+        gr.Image(type="pil", label="Processed Image")
+    ],
+    title="Bank Statement OCR",
+    # description="Upload an image containing text to perform OCR and see the detected text and image."
+    flagging_options=["blurry", "incorrect", "other"],)
+
+iface.launch(debug=True)