Update app.py

app.py

@@ -34,7 +34,6 @@ translation_model = MarianMTModel.from_pretrained('Helsinki-NLP/opus-mt-hi-en')
 
 # Define a function for keyword highlighting
 def highlight_keywords(text, keyword):
-    # Escape keyword for regex to avoid issues with special characters
     pattern = re.compile(re.escape(keyword), re.IGNORECASE)
     highlighted_text = pattern.sub(lambda match: f"**{match.group(0)}**", text)
     return highlighted_text
@@ -59,11 +58,11 @@ if image_file is not None:
     if st.button("Run OCR"):
         # Use GOT-OCR2 model for plain text OCR (structured documents)
         with torch.no_grad():
-            res_plain = model.chat(tokenizer, temp_file_path, ocr_type='ocr')  # Removed device parameter
+            res_plain = model.chat(tokenizer, temp_file_path, ocr_type='ocr')
 
         # Perform formatted text OCR
         with torch.no_grad():
-            res_format = model.chat(tokenizer, temp_file_path, ocr_type='format')  # Removed device parameter
+            res_format = model.chat(tokenizer, temp_file_path, ocr_type='format')
 
         # Use EasyOCR for both English and Hindi text recognition
         result_easyocr = reader.readtext(temp_file_path, detail=0)
@@ -94,13 +93,13 @@ if image_file is not None:
 
         # Additional OCR types using GOT-OCR2
         with torch.no_grad():
-            res_fine_grained = model.chat(tokenizer, temp_file_path, ocr_type='ocr', ocr_box='')  # Removed device parameter
+            res_fine_grained = model.chat(tokenizer, temp_file_path, ocr_type='ocr', ocr_box='')
         st.subheader("Fine-Grained OCR Results:")
         st.write(res_fine_grained)
 
         # Render formatted OCR to HTML
         with torch.no_grad():
-            res_render = model.chat(tokenizer, temp_file_path, ocr_type='format', render=True, save_render_file='./demo.html')  # Removed device parameter
+            res_render = model.chat(tokenizer, temp_file_path, ocr_type='format', render=True, save_render_file='./demo.html')
         st.subheader("Rendered OCR Results (HTML):")
         st.write(res_render)
 
@@ -109,11 +108,8 @@ if image_file is not None:
 
         if keyword:
             st.subheader("Search Results:")
-            # Highlight the matching sections in the extracted text
             highlighted_text = highlight_keywords(extracted_text, keyword)
             st.markdown(highlighted_text)
 
         # Clean up the temporary file after use
         os.remove(temp_file_path)
-
-# Note: No need for if __name__ == "__main__": st.run()