def translate(text):
"""Attempt to translate text using food_words and then the CEDICT dictionary."""
start = dt.datetime.now()
words = find_words(text)
start = time_elapsed("Find words", start)
results = check_words(words)
start = time_elapsed("Check words", start)
return results
def translate(text):
"""Attempt to translate text using food_words and then the CEDICT dictionary."""
start = dt.datetime.now()
words = find_words(text)
start = time_elapsed("Find words", start)
results = check_words(words)
start = time_elapsed("Check words", start)
return results
def search(text):
"""Takes a string and returns dish information or translation"""
# timing information, can delete later
start = datetime.datetime.now()
# Returns search data for a particular query.
results = search_dish_name(text)
time_elapsed("Search and translate", start)
return json.dumps(results)
def search(text):
"""Takes a string and returns dish information or translation"""
# timing information, can delete later
start = datetime.datetime.now()
# Returns search data for a particular query.
results = search_dish_name(text)
time_elapsed("Search and translate", start)
return json.dumps(results)
def preprocess_image(path):
"""Do standard preprocessing on image before sending to Tesseract."""
start = dt.datetime.now()
# Steps using PIL
im = Image.open(path)
im = smooth_and_grayscale(im)
start = time_elapsed("Smoothing", start)
im = binarize(im)
start = time_elapsed("Binarization", start)
im.save(path)
def preprocess_image(path):
"""Do standard preprocessing on image before sending to Tesseract."""
start = dt.datetime.now()
# Steps using PIL
im = Image.open(path)
im = smooth_and_grayscale(im)
start = time_elapsed("Smoothing", start)
im = binarize(im)
start = time_elapsed("Binarization", start)
im.save(path)
def upload():
"""Handler for uploading an image, processing and sending to Tesseract."""
if request.data:
file = request.data
now = datetime.datetime.utcnow()
# Create a unique filename for the image.
filename = now.strftime('%Y%m%d%M%S') + '.png'
image_path = os.path.join(app.config['UPLOAD_FOLDER'], filename)
# write the image data to the file
with open(image_path, 'wb') as f:
f.write(file)
# timing information for performance, can delete later
start = datetime.datetime.now()
start = time_elapsed("Writing file", start)
# do some preprocessing on the image to optimize it for Tesseract
preprocess_image(image_path)
start = time_elapsed("Preprocessing", start)
# run the image through tesseract and extract text
text = image_file_to_string(image_path,
lang="chi_sim",
graceful_errors=True)
text = text.strip()
start = time_elapsed("Tesseract", start)
if not text:
# if Tesseract returns nothing, do some additional processing to see if results improve.
smooth_and_thin_image(image_path)
# run through tesseract again
text = image_file_to_string(image_path,
lang=LANG,
graceful_errors=True)
text = text.strip()
start = time_elapsed("Tesseract", start)
if not text:
# if still no text from Tesseract, send an error to the client
error_data = {"error": "No results found. Please try again."}
return json.dumps(error_data)
# if text was received, redirect to search
return redirect(url_for("search", text=text))
def upload():
"""Handler for uploading an image, processing and sending to Tesseract."""
if request.data:
file = request.data
now = datetime.datetime.utcnow()
# Create a unique filename for the image.
filename = now.strftime('%Y%m%d%M%S') + '.png'
image_path = os.path.join(app.config['UPLOAD_FOLDER'], filename)
# write the image data to the file
with open(image_path, 'wb') as f:
f.write(file)
# timing information for performance, can delete later
start = datetime.datetime.now()
start = time_elapsed("Writing file", start)
# do some preprocessing on the image to optimize it for Tesseract
preprocess_image(image_path)
start = time_elapsed("Preprocessing", start)
# run the image through tesseract and extract text
text = image_file_to_string(image_path, lang="chi_sim", graceful_errors=True)
text = text.strip()
start = time_elapsed("Tesseract", start)
if not text:
# if Tesseract returns nothing, do some additional processing to see if results improve.
smooth_and_thin_image(image_path)
# run through tesseract again
text = image_file_to_string(image_path, lang=LANG, graceful_errors=True)
text = text.strip()
start = time_elapsed("Tesseract", start)
if not text:
# if still no text from Tesseract, send an error to the client
error_data = {"error": "No results found. Please try again."}
return json.dumps(error_data)
# if text was received, redirect to search
return redirect(url_for("search", text=text))
def search_dish_name(text):
"""Searches for text in the dishes database. If not found, translates text and
looks for similar dishes in database. Returns JSON data for dish or search results."""
# timing information, can delete later.
start = dt.datetime.now()
results = {}
if type(text) != unicode:
text = text.decode('utf-8')
if len(text) > 10:
# Most dish names are 3-5 characters.
# If Tesseract returned more than 10 characters, something probably went wrong.
print "Input text is too long."
return None
else:
# Find a matching dish, if it exists.
match = Dish.find_match(text)
if match:
# If result is found, return JSON representation of dish.
results = match.get_json()
start = time_elapsed("Dish lookup", start)
else:
# If no dish is found, return translation data and similar dishes, if they exist.
translation = translate(text)
start = time_elapsed("Translation", start)
results['translation'] = translation
# Find similar dishes and add to results.
if len(text) > 1:
similar_dishes = Dish.find_similar(text)
start = time_elapsed("Similar dish lookup", start)
similar_json = []
for similar_dish in similar_dishes:
dish_data = similar_dish.get_json_min()
similar_json.append(dish_data)
if similar_json != []:
results['similar'] = similar_json
return results
def search_dish_name(text):
"""Searches for text in the dishes database. If not found, translates text and
looks for similar dishes in database. Returns JSON data for dish or search results."""
# timing information, can delete later.
start = dt.datetime.now()
results = {}
if type(text) != unicode:
text = text.decode('utf-8')
if len(text) > 10:
# Most dish names are 3-5 characters.
# If Tesseract returned more than 10 characters, something probably went wrong.
print "Input text is too long."
return None
else:
# Find a matching dish, if it exists.
match = Dish.find_match(text)
if match:
# If result is found, return JSON representation of dish.
results = match.get_json()
start = time_elapsed("Dish lookup", start)
else:
# If no dish is found, return translation data and similar dishes, if they exist.
translation = translate(text)
start = time_elapsed("Translation", start)
results['translation'] = translation
# Find similar dishes and add to results.
if len(text) > 1:
similar_dishes = Dish.find_similar(text)
start = time_elapsed("Similar dish lookup", start)
similar_json = []
for similar_dish in similar_dishes:
dish_data = similar_dish.get_json_min()
similar_json.append(dish_data)
if similar_json != []:
results['similar'] = similar_json
return results
def smooth_and_thin_image(path):
"""Extra preprocessing, including smoothing, acute angle emphasis and thinning."""
start = dt.datetime.now()
# Steps using PIL
im = Image.open(path)
# Steps using scikit-image
pix = im_to_trutharray(im)
# Stentiford preprocessing for image thinning.
smooth_and_emphasize_angles(pix)
start = time_elapsed("Stentiford preprocessing", start)
# Thinning
pix = thin_image(pix)
start = time_elapsed("Thinning", start)
im = trutharray_to_im(pix)
im.save(path)
def smooth_and_thin_image(path):
"""Extra preprocessing, including smoothing, acute angle emphasis and thinning."""
start = dt.datetime.now()
# Steps using PIL
im = Image.open(path)
# Steps using scikit-image
pix = im_to_trutharray(im)
# Stentiford preprocessing for image thinning.
smooth_and_emphasize_angles(pix)
start = time_elapsed("Stentiford preprocessing", start)
# Thinning
pix = thin_image(pix)
start = time_elapsed("Thinning", start)
im = trutharray_to_im(pix)
im.save(path)
