Update app.py (#1)

- Update app.py (3ea5660c1221889b32e153b1f56482e986ab4b6d)
- Update requirements.txt (b5c6ceff2ba8b78c1700265eee54a4029512451b)

app.py

@@ -4,6 +4,9 @@ from PIL import Image
 import torch
 import spaces
 import json
+import re
+from langdetect import detect, LangDetectException
+from googletrans import Translator
 
 # Load the processor and model
 processor = AutoProcessor.from_pretrained(
@@ -69,6 +72,95 @@ def wrap_json_in_markdown(text):
     result.append(text)  # Append any remaining text
     return ''.join(result)
 
+def decode_unicode_sequences(unicode_seq):
+    """
+    Decodes a sequence of Unicode escape sequences (e.g., \\u4F60\\u597D) to actual characters.
+
+    Args:
+        unicode_seq (str): A string containing Unicode escape sequences.
+
+    Returns:
+        str: The decoded Unicode string.
+    """
+    # Regular expression to find \uXXXX
+    unicode_escape_pattern = re.compile(r'\\u([0-9a-fA-F]{4})')
+    
+    # Function to replace each \uXXXX with the corresponding character
+    def replace_match(match):
+        hex_value = match.group(1)
+        return chr(int(hex_value, 16))
+    
+    # Decode all \uXXXX sequences
+    decoded = unicode_escape_pattern.sub(replace_match, unicode_seq)
+    return decoded
+
+def is_mandarin(text):
+    """
+    Detects if the given text is in Mandarin.
+
+    Args:
+        text (str): The text to check.
+
+    Returns:
+        bool: True if the text is detected as Mandarin, False otherwise.
+    """
+    try:
+        lang = detect(text)
+        return lang == 'zh-cn' or lang == 'zh-tw' or lang == 'zh'
+    except LangDetectException:
+        return False
+
+def translate_to_english(text, translator):
+    """
+    Translates the given Mandarin text to English.
+
+    Args:
+        text (str): The Mandarin text to translate.
+        translator (Translator): An instance of googletrans Translator.
+
+    Returns:
+        str: The translated English text.
+    """
+    try:
+        translation = translator.translate(text, src='zh-cn', dest='en')
+        return translation.text
+    except Exception as e:
+        print(f"Translation error: {e}")
+        return text  # Return the original text if translation fails
+
+def process_text_for_mandarin_unicode(input_string):
+    """
+    Processes the input string to find Unicode escape sequences representing Mandarin words,
+    translates them to English, and replaces them accordingly.
+
+    Args:
+        input_string (str): The original string containing Unicode escape sequences.
+
+    Returns:
+        str: The processed string with translations where applicable.
+    """
+    # Initialize the translator
+    translator = Translator()
+    
+    # Regular expression to find groups of consecutive \uXXXX sequences
+    unicode_word_pattern = re.compile(r'(?:\\u[0-9a-fA-F]{4})+')
+    
+    # Function to process each matched Unicode word
+    def process_match(match):
+        unicode_seq = match.group(0)
+        decoded_word = decode_unicode_sequences(unicode_seq)
+        
+        if is_mandarin(decoded_word):
+            translated = translate_to_english(decoded_word, translator)
+            return f"{translated} ({decoded_word})"
+        else:
+            # If not Mandarin, return the original sequence
+            return unicode_seq
+    
+    # Substitute all matched Unicode words with their translations if applicable
+    processed_string = unicode_word_pattern.sub(process_match, input_string)
+    return processed_string
+
 @spaces.GPU()
 def process_image_and_text(image, text):
     # Process the image and text
@@ -91,8 +183,9 @@ def process_image_and_text(image, text):
     generated_tokens = output[0, inputs['input_ids'].size(1):]
     generated_text = processor.tokenizer.decode(generated_tokens, skip_special_tokens=True)
     generated_text_w_json_wrapper = wrap_json_in_markdown(generated_text)
+    generated_text_w_unicode_mdn = process_text_for_mandarin_unicode(generated_text_w_json_wrapper)
     
-    return generated_text_w_json_wrapper
+    return generated_text_w_unicode_mdn
 
 def chatbot(image, text, history):
     if image is None:

requirements.txt

@@ -4,4 +4,6 @@ Pillow
 torchvision 
 einops
 accelerate
-tensorflow
+tensorflow
+langdetect
+googletrans