def My_Object_Detection(url, filename):
SUBSCRIPTION_KEY = os.getenv('Object_Detection_KEY')
ENDPOINT = os.getenv('Object_Detection_ENDPOINT')
CV_CLIENT = ComputerVisionClient(
ENDPOINT, CognitiveServicesCredentials(SUBSCRIPTION_KEY))
IMGUR_CONFIG = {
"client_id": os.getenv('IMGUR_Client_ID'),
"client_secret": os.getenv('IMGUR_Client_Secret'),
"access_token": os.getenv('Postman_Access_Token'),
"refresh_token": os.getenv('Postman_Refresh_token')
}
IMGUR_CLIENT = Imgur(config=IMGUR_CONFIG)
img = Image.open(filename)
draw = ImageDraw.Draw(img)
font_size = int(5e-2 * img.size[1])
fnt = ImageFont.truetype("./font/TaipeiSansTCBeta-Regular.ttf",
size=font_size)
object_detection = CV_CLIENT.detect_objects(url) # create detection_object
if len(object_detection.objects) > 0:
for obj in object_detection.objects:
left = obj.rectangle.x
top = obj.rectangle.y
right = obj.rectangle.x + obj.rectangle.w
bot = obj.rectangle.y + obj.rectangle.h
name = obj.object_property # prediction of object
confidence = obj.confidence
draw.rectangle([left, top, right, bot],
outline=(255, 0, 0),
width=3)
draw.text(
[left, top + font_size],
"{0} {1:0.1f}".format(name, confidence * 100),
fill=(255, 0, 0),
font=fnt,
)
img.save(filename)
image = IMGUR_CLIENT.image_upload(filename, "title", "description")
link = image["response"]["data"]["link"]
os.remove(filename)
return link
def detectObjects(containerName, blobName):
"""
Description: This function detects objects, using Microsofts Azure Computer Vision Algorithm, in an image retrieved from an Azure Blob Storage, via its URL.
Source: The code has been written with the help of the Azure Computer Vision Documentation (https://docs.microsoft.com/en-us/python/api/azure-cognitiveservices-vision-computervision/azure.cognitiveservices.vision.computervision.operations.computervisionclientoperationsmixin?view=azure-python#detect-objects-in-stream-image--custom-headers-none--raw-false--callback-none----operation-config-)
Input: (containerName -> string), (blobName -> string)
Output: List -> [[object.object_property, object.confidence, object.rectangle.x, object.rectangle.w, object.rectangle.y, object.rectangle.h], ...]
"""
print("\n\n===== Detect Objects=====\n")
#Retriving the probability threshold to recognise an object
result_probThreshold = dbc.retrieve_probThreshold("computerVisionObjects")
probThreshold = float(result_probThreshold[0])
# Client Authentication
computervision_client = ComputerVisionClient(
CV_endpoint, CognitiveServicesCredentials(CV_subscription_key))
# Get URL image with different objects
remote_image_url_objects = hbl.getBlobURI(containerName, blobName)
# Call API with URL
detect_objects_results_remote = computervision_client.detect_objects(
remote_image_url_objects)
# Detect objects in an image and store results in nested resultList of form: [[object.object_property, object.confidence], [object.object_property, object.confidence], ...]
resultList = []
print("Detecting objects in remote image:")
if len(detect_objects_results_remote.objects) == 0:
print("No objects detected.")
else:
for object in detect_objects_results_remote.objects:
if (object.confidence > probThreshold):
objectList = [
object.object_property, object.confidence,
object.rectangle.x, object.rectangle.w, object.rectangle.y,
object.rectangle.h
]
resultList.append(objectList)
return resultList
# Create client.
client = ComputerVisionClient(endpoint, credentials)
# Check the URL supplied or path exists and is an image.
# Send provided image (url or path) to azure to extract text.
# ----------------------------------------------------------------------
# URL or path
# ----------------------------------------------------------------------
path = args.path
# ----------------------------------------------------------------------
# Objects
# ----------------------------------------------------------------------
if is_url(path):
analysis = client.detect_objects(path)
else:
path = os.path.join(get_cmd_cwd(), path)
with open(path, 'rb') as fstream:
analysis = client.detect_objects_in_stream(fstream)
for object in analysis.objects:
print(f"{object.rectangle.x} {object.rectangle.y} " +
f"{object.rectangle.x + object.rectangle.w} " +
f"{object.rectangle.y + object.rectangle.h}")
if len(local_image_objects.objects) == 0:
print("No objects detected.")
else:
for object in local_image_objects.objects:
print("object at location {}, {}, {}, {}".format( \
object.rectangle.x, object.rectangle.x + object.rectangle.w, \
object.rectangle.y, object.rectangle.y + object.rectangle.h))
# END - Detect objects in a local image
# Detect objects in a remote image by:
# 1. Opening the binary file for reading.
# 2. Calling the Computer Vision service's detect_objects with the:
# - image
# - features to extract
# 3. Displaying the location of the objects.
remote_image_objects = computervision_client.detect_objects(remote_image_url)
print("\nDetecting objects in remote image:")
if len(remote_image_objects.objects) == 0:
print("No objects detected.")
else:
for object in remote_image_objects.objects:
print("object at location {}, {}, {}, {}".format( \
object.rectangle.x, object.rectangle.x + object.rectangle.w, \
object.rectangle.y, object.rectangle.y + object.rectangle.h))
# END - Detect objects in a remote image
# Detect brands in a local image by:
# 1. Opening the binary file for reading.
# 2. Calling the Computer Vision service's analyze_image_in_stream with the:
# - image
object.rectangle.y, object.rectangle.y + object.rectangle.h))
print()
'''
END - Detect Objects - local
'''
# <snippet_objects>
'''
Detect Objects - remote
This example detects different kinds of objects with bounding boxes in a remote image.
'''
print("===== Detect Objects - remote =====")
# Get URL image with different objects
remote_image_url_objects = "https://raw.githubusercontent.com/Azure-Samples/cognitive-services-sample-data-files/master/ComputerVision/Images/objects.jpg"
# Call API with URL
detect_objects_results_remote = computervision_client.detect_objects(remote_image_url_objects)
# Print detected objects results with bounding boxes
print("Detecting objects in remote image:")
if len(detect_objects_results_remote.objects) == 0:
print("No objects detected.")
else:
for object in detect_objects_results_remote.objects:
print("object at location {}, {}, {}, {}".format( \
object.rectangle.x, object.rectangle.x + object.rectangle.w, \
object.rectangle.y, object.rectangle.y + object.rectangle.h))
# </snippet_objects>
print()
'''
END - Detect Objects - remote
'''
def getNewDataFromMS(self, cleanedData):
data_algorithm = []
self.Aux.deleteDataFromFolder('./Data/Response/')
computervision_client = ComputerVisionClient(
self.VISION_ENDPOINT,
CognitiveServicesCredentials(self.VISION_KEY))
ta_credential = AzureKeyCredential(self.TEXT_KEY)
text_analytics_client = TextAnalyticsClient(
endpoint=self.TEXT_ENDPOINT, credential=ta_credential)
for row in cleanedData:
try:
# PIC ANALYTICS
remote_image_url = row['algorithm']['photo']
remote_image_url_ed = row['algorithm']['photo_ed']
description_results = computervision_client.describe_image(
remote_image_url)
if len(description_results.captions) > 0:
tags = description_results.tags
row['results']['TagsInPic'] = tags
row['results']['NumOfObjectsInPic'] = len(tags)
if len(tags) <= 5:
row['results']['CLASS_fewObjects'] = True
elif len(tags) >= 20:
row['results']['CLASS_manyObjects'] = True
else:
row['results']['CLASS_normalObjects'] = True
for caption in description_results.captions:
confidence = caption.confidence * 100
if confidence > 50:
row['results']['hasContent'] = 'yes'
row['results']['content'].append(caption.text)
else:
row['results']['hasContent'] = 'unsure'
break
# get the picture category
remote_image_features = ["categories"]
categorize_results_remote = computervision_client.analyze_image(
remote_image_url, remote_image_features)
if len(categorize_results_remote.categories) > 0:
for category in categorize_results_remote.categories:
if category.score * 100 > 50:
row['results']['imageCategory'].append(
category.name)
else:
row['results']['imageCategory'].append(
'unsure')
# get all objects in picture
detect_objects_results_remote = computervision_client.detect_objects(
remote_image_url)
for objects in detect_objects_results_remote.objects:
if objects.object_property == 'person' and objects.confidence * 100 > 50:
row['results']['hasHuman'] = True
# check if a face of the person is visible
remote_image_features = ["faces"]
detect_faces_results_remote = computervision_client.analyze_image(
remote_image_url, remote_image_features)
if len(detect_faces_results_remote.faces) > 0:
row['results']['hasFace'] = True
# Color scheme
remote_image_features = ["color"]
detect_color_results_remote = computervision_client.analyze_image(
remote_image_url, remote_image_features)
detect_color_results_remote_ed = computervision_client.analyze_image(
remote_image_url_ed, remote_image_features)
picColor = detect_color_results_remote
if not picColor.color.is_bw_img:
row['results']['hasColor'] = True
background = picColor.color.dominant_color_background
row['colors']['background'] = background
foreground = picColor.color.dominant_color_foreground
row['colors']['foreground'] = foreground
dominantColors = detect_color_results_remote_ed.color.dominant_colors
row['colors']['dominantColors'] = dominantColors
accentColor = picColor.color.accent_color
row['colors']['accentColor'] = accentColor
if background == 'Black' and foreground == 'Black':
row['results']['isBright'] = False
if len(dominantColors) > 2:
row['results']['hasManyDomColors'] = True
answer = self.Aux.getHue(accentColor)
hue = answer[1]
row['colors']['hue'] = hue
warmHue = answer[0]
if warmHue:
row['results']['hasWarmHueAccent'] = True
# TEXT ANALYTICS
title = [row['algorithm']['title']]
# sentiment and length TITLE
sentimentTitle = self.getLengthSentiment(
title, text_analytics_client)
row['results']['lengthOfTitle'] = sentimentTitle[0]
if sentimentTitle[0] >= 47:
row['results']['CLASS_longTitle'] = True
elif sentimentTitle[0] <= 22:
row['results']['CLASS_shortTitle'] = True
else:
row['results']['CLASS_normalTitle'] = True
row['results']['sentimentTitle'] = sentimentTitle[1]
if sentimentTitle[1] == 'positive':
row['results']['CLASS_positiveTitle'] = True
elif sentimentTitle[1] == 'neutral':
row['results']['CLASS_neutralTitle'] = True
else:
row['results']['CLASS_negativeTitle'] = True
row['results']['sentiScoresTitle'] = sentimentTitle[
2] # pos neu neg share
# get Key Phrases in TITLE
phrasesTitle = self.getPhrases(title, text_analytics_client)
keyPhrasesTitle = []
for phrase in phrasesTitle:
phrase.replace('-', ' ')
wordList = re.sub("[^\w]", " ", phrase).split()
keyPhrasesTitle.append(wordList)
flattenedKeyPhrasesTitle = list(
self.Aux.flatten(keyPhrasesTitle))
row['results']['keyPhrasesTitle'] = flattenedKeyPhrasesTitle
# analyze TEXT
text = [row['algorithm']['text']]
# sentiment and length TEXT
sentimentText = self.getLengthSentiment(
text, text_analytics_client)
row['results']['lengthOfText'] = sentimentText[0]
if sentimentText[0] >= 131:
row['results']['CLASS_longText'] = True
elif sentimentText[0] <= 100:
row['results']['CLASS_shortText'] = True
else:
row['results']['CLASS_normalText'] = True
row['results']['sentimentText'] = sentimentText[1]
if sentimentText[1] == 'positive':
row['results']['CLASS_positiveText'] = True
elif sentimentText[1] == 'neutral':
row['results']['CLASS_neutralText'] = True
else:
row['results']['CLASS_negativeText'] = True
row['results']['sentiScoresText'] = sentimentText[2]
# get Key Phrases TEXT
phrasesText = self.getPhrases(text, text_analytics_client)
keyPhrasesText = []
for phrase in phrasesText:
phrase.replace('-', ' ')
wordList = re.sub("[^\w]", " ", phrase).split()
keyPhrasesText.append(wordList)
flattenedKeyPhrasesText = list(
self.Aux.flatten(keyPhrasesText))
row['results']['keyPhrasesText'] = flattenedKeyPhrasesText
# analyze TITLE TEXT and Picture
picTags = row['results']['TagsInPic']
phrases = flattenedKeyPhrasesText + flattenedKeyPhrasesTitle
matchPic = self.Aux.textMatch(phrases, picTags)
row['results']['TextMatchPic'] = matchPic[1]
# analyze creator and TITLE TEXT
creator = row['algorithm']['creator']
matchCreator = self.Aux.textMatch(phrases, creator)
row['results']['CreatorMatchTitle'] = matchCreator[1]
# analyze OCR in picture
picUrl = row['algorithm']['photo'].lstrip("'")
OCRTags = self.getOCRTags(picUrl)
if len(OCRTags) > 0:
row['results']['OCRTags'] = OCRTags
TextMatchOCR = self.Aux.textMatch(phrases, OCRTags)
row['results']['OCRMatches'] = TextMatchOCR[0]
row['results']['TextMatchPicOCR'] = TextMatchOCR[1]
# check HYPOTHESIS
if row['results']['hasHuman'] and row['results']['hasColor'] and row['results']['isBright'] and \
not row['results']['CLASS_negativeTitle'] and \
row['results']['CLASS_positiveText'] and row['results']['hasWarmHueAccent']:
row['results']['H1_Emotion'] = True
if not row['results']['CLASS_manyObjects'] and not row['results']['CLASS_longTitle'] and \
not row['results']['CLASS_longText'] and \
row['results']['CLASS_neutralText'] and not row['results']['CLASS_negativeTitle']:
row['results']['H2_ClearMassage'] = True
if row['results']['CreatorMatchTitle'] and row['results'][
'TextMatchPicOCR']:
row['results']['H3_Trust'] = True
data_algorithm.append(row)
except Exception as E:
print('Project cannot be analyzed by Algrorithm ' + str(E))
return data_algorithm
