Update app.py

app.py

@@ -1,59 +1,78 @@
-import gradio as gr
-import easyocr
-from deep_translator import GoogleTranslator
-from PIL import Image, ImageDraw, ImageFont
+import cv2
+from PIL import Image
 import numpy as np
+import pytesseract
+from googletrans import Translator
 
-# Initialize reader with allowed lang_list only
-reader = easyocr.Reader(['en', 'hi', 'mr', 'ne'])
+def overlay_translation(image_path, src_lang='hi', dest_lang='en'):
+    # Load image
+    image = cv2.imread(image_path)
+    image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
 
-# Font path — works on HuggingFace
-FONT_PATH = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"
+    # OCR data with boxes
+    data = pytesseract.image_to_data(image_rgb, lang=src_lang, output_type=pytesseract.Output.DICT)
 
-def process_image(image, target_lang):
-    img_array = np.array(image)
+    translator = Translator()
+    translated_image = image_rgb.copy()
 
-    # Run OCR safely
-    results = reader.readtext(img_array)
+    for i in range(len(data['text'])):
+        text = data['text'][i].strip()
+        if text != "":
+            x, y, w, h = data['left'][i], data['top'][i], data['width'][i], data['height'][i]
 
-    if not results:
-        return image, "No text detected"
+            try:
+                translated_text = translator.translate(text, src=src_lang, dest=dest_lang).text
+            except:
+                translated_text = text  # fallback
+
+            # White rectangle behind text for better visibility
+            cv2.rectangle(translated_image, (x, y), (x+w, y+h), (255,255,255), -1)
+
+            # Add translated text (in black, centered)
+            font_scale = h / 35
+            font_thickness = 2
+            cv2.putText(translated_image, translated_text, (x, y + h - 5), 
+                        cv2.FONT_HERSHEY_SIMPLEX, font_scale, (0,0,0), font_thickness, cv2.LINE_AA)
+
+    # Save or return the translated image
+    translated_image_pil = Image.fromarray(translated_image)
+    return translated_image_pil
+import cv2
+from PIL import Image
+import numpy as np
+import pytesseract
+from googletrans import Translator
 
-    final_translations = []
-    img_pil = image.convert("RGB")
-    draw = ImageDraw.Draw(img_pil)
-    font = ImageFont.truetype(FONT_PATH, size=20)
+def overlay_translation(image_path, src_lang='hi', dest_lang='en'):
+    # Load image
+    image = cv2.imread(image_path)
+    image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+
+    # OCR data with boxes
+    data = pytesseract.image_to_data(image_rgb, lang=src_lang, output_type=pytesseract.Output.DICT)
+
+    translator = Translator()
+    translated_image = image_rgb.copy()
+
+    for i in range(len(data['text'])):
+        text = data['text'][i].strip()
+        if text != "":
+            x, y, w, h = data['left'][i], data['top'][i], data['width'][i], data['height'][i]
 
-    for (bbox, text, prob) in results:
-        if text.strip():
             try:
-                translated = GoogleTranslator(source='auto', target=target_lang).translate(text)
+                translated_text = translator.translate(text, src=src_lang, dest=dest_lang).text
             except:
-                translated = "[Translation Error]"
-
-            final_translations.append(translated)
-
-            # Draw translation at top-left of bounding box
-            top_left = bbox[0]
-            draw.text((top_left[0], top_left[1] - 20), translated, fill=(255, 0, 0), font=font)
-
-    full_text = "\n".join(final_translations)
-    return img_pil, full_text
-
-# Gradio UI
-iface = gr.Interface(
-    fn=process_image,
-    inputs=[
-        gr.Image(type="pil", label="Upload Image"),
-        gr.Textbox(label="Translate To (hi / en / mr / ne)")
-    ],
-    outputs=[
-        gr.Image(type="pil", label="Image with Translated Overlay"),
-        gr.Textbox(label="Translated Text Output")
-    ],
-    title="Image Translator with Overlay",
-    description="Extract & translate text from an image (OCR + translation overlay)"
-)
-
-if __name__ == "__main__":
-    iface.launch()
+                translated_text = text  # fallback
+
+            # White rectangle behind text for better visibility
+            cv2.rectangle(translated_image, (x, y), (x+w, y+h), (255,255,255), -1)
+
+            # Add translated text (in black, centered)
+            font_scale = h / 35
+            font_thickness = 2
+            cv2.putText(translated_image, translated_text, (x, y + h - 5), 
+                        cv2.FONT_HERSHEY_SIMPLEX, font_scale, (0,0,0), font_thickness, cv2.LINE_AA)
+
+    # Save or return the translated image
+    translated_image_pil = Image.fromarray(translated_image)
+    return translated_image_pil