Quickstart: Computer Vision client library for Python

The Computer Vision service provides developers with access to advanced algorithms for processing images and returning information. Computer Vision algorithms analyze the content of an image in different ways, depending on the visual features you're interested in.

Use the Computer Vision client library for Python to:

  • Analyze an image for tags, text description, faces, adult content, and more.
  • Recognize printed and handwritten text with the Batch Read API.


The scenarios in this quickstart use remote image URLs. For sample code that does the same operations on local images, see the code on GitHub.

Reference documentation | Library source code | Package (PiPy) | Samples


Setting up

Create a Computer Vision Azure resource

Azure Cognitive Services are represented by Azure resources that you subscribe to. Create a resource for Computer Vision using the Azure portal or Azure CLI on your local machine. You can also:

After you get a key from your trial subscription or resource, create environment variables for the key and endpoint URL, named COMPUTER_VISION_SUBSCRIPTION_KEY and COMPUTER_VISION_ENDPOINT, respectively.

Create a new Python application

Create a new Python script—quickstart-file.py, for example. Then open it in your preferred editor or IDE and import the following libraries.

from azure.cognitiveservices.vision.computervision import ComputerVisionClient
from azure.cognitiveservices.vision.computervision.models import TextOperationStatusCodes
from azure.cognitiveservices.vision.computervision.models import TextRecognitionMode
from azure.cognitiveservices.vision.computervision.models import VisualFeatureTypes
from msrest.authentication import CognitiveServicesCredentials

from array import array
import os
from PIL import Image
import sys
import time

Then, create variables for your resource's Azure endpoint and key.

# Add your Computer Vision subscription key to your environment variables.
    subscription_key = os.environ['COMPUTER_VISION_SUBSCRIPTION_KEY']
    print("\nSet the COMPUTER_VISION_SUBSCRIPTION_KEY environment variable.\n**Restart your shell or IDE for changes to take effect.**")
# Add your Computer Vision endpoint to your environment variables.
if 'COMPUTER_VISION_ENDPOINT' in os.environ:
    endpoint = os.environ['COMPUTER_VISION_ENDPOINT']
    print("\nSet the COMPUTER_VISION_ENDPOINT environment variable.\n**Restart your shell or IDE for changes to take effect.**")


If you created the environment variable after you launched the application, you will need to close and reopen the editor, IDE, or shell running it to access the variable.

Install the client library

You can install the client library with:

pip install --upgrade azure-cognitiveservices-vision-computervision

Object model

The following classes and interfaces handle some of the major features of the Computer Vision Python SDK.

Name Description
ComputerVisionClientOperationsMixin This class directly handles all of the image operations, such as image analysis, text detection, and thumbnail generation.
ComputerVisionClient This class is needed for all Computer Vision functionality. You instantiate it with your subscription information, and you use it to produce instances of other classes. It implements ComputerVisionClientOperationsMixin.
VisualFeatureTypes This enum defines the different types of image analysis that can be done in a standard Analyze operation. You specify a set of VisualFeatureTypes values depending on your needs.

Code examples

These code snippets show you how to do the following tasks with the Computer Vision client library for Python:

Authenticate the client


This quickstart assumes you've created an environment variable for your Computer Vision key, named COMPUTER_VISION_SUBSCRIPTION_KEY.

Instantiate a client with your endpoint and key. Create a CognitiveServicesCredentials object with your key, and use it with your endpoint to create a ComputerVisionClient object.

computervision_client = ComputerVisionClient(endpoint, CognitiveServicesCredentials(subscription_key))

Analyze an image

Save a reference to the URL of an image you want to analyze.

remote_image_url = "https://raw.githubusercontent.com/Azure-Samples/cognitive-services-sample-data-files/master/ComputerVision/Images/landmark.jpg"

Get image description

The following code gets the list of generated captions for the image. See Describe images for more details.

Describe an Image - remote
This example describes the contents of an image with the confidence score.
print("===== Describe an image - remote =====")
# Call API
description_results = computervision_client.describe_image(remote_image_url )

# Get the captions (descriptions) from the response, with confidence level
print("Description of remote image: ")
if (len(description_results.captions) == 0):
    print("No description detected.")
    for caption in description_results.captions:
        print("'{}' with confidence {:.2f}%".format(caption.text, caption.confidence * 100))

Get image category

The following code gets the detected category of the image. See Categorize images for more details.

Categorize an Image - remote
This example extracts (general) categories from a remote image with a confidence score.
print("===== Categorize an image - remote =====")
# Select the visual feature(s) you want.
remote_image_features = ["categories"]
# Call API with URL and features
categorize_results_remote = computervision_client.analyze_image(remote_image_url , remote_image_features)

# Print results with confidence score
print("Categories from remote image: ")
if (len(categorize_results_remote.categories) == 0):
    print("No categories detected.")
    for category in categorize_results_remote.categories:
        print("'{}' with confidence {:.2f}%".format(category.name, category.score * 100))

Get image tags

The following code gets the set of detected tags in the image. See Content tags for more details.

Tag an Image - remote
This example returns a tag (key word) for each thing in the image.
print("===== Tag an image - remote =====")
# Call API with remote image
tags_result_remote = computervision_client.tag_image(remote_image_url )

# Print results with confidence score
print("Tags in the remote image: ")
if (len(tags_result_remote.tags) == 0):
    print("No tags detected.")
    for tag in tags_result_remote.tags:
        print("'{}' with confidence {:.2f}%".format(tag.name, tag.confidence * 100))

Detect objects

The following code detects common objects in the image and prints them to the console. See Object detection for more details.

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.")
    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))

Detect brands

The following code detects corporate brands and logos in the image and prints them to the console. See Brand detection for more details.

Detect Brands - remote
This example detects common brands like logos and puts a bounding box around them.
print("===== Detect Brands - remote =====")
# Get a URL with a brand logo
remote_image_url = "https://docs.microsoft.com/en-us/azure/cognitive-services/computer-vision/images/gray-shirt-logo.jpg"
# Select the visual feature(s) you want
remote_image_features = ["brands"]
# Call API with URL and features
detect_brands_results_remote = computervision_client.analyze_image(remote_image_url, remote_image_features)

print("Detecting brands in remote image: ")
if len(detect_brands_results_remote.brands) == 0:
    print("No brands detected.")
    for brand in detect_brands_results_remote.brands:
        print("'{}' brand detected with confidence {:.1f}% at location {}, {}, {}, {}".format( \
        brand.name, brand.confidence * 100, brand.rectangle.x, brand.rectangle.x + brand.rectangle.w, \
        brand.rectangle.y, brand.rectangle.y + brand.rectangle.h))

Detect faces

The following code returns the detected faces in the image with their rectangle coordinates and select face attributes. See Face detection for more details.

Detect Faces - remote
This example detects faces in a remote image, gets their gender and age, 
and marks them with a bounding box.
print("===== Detect Faces - remote =====")
# Get an image with faces
remote_image_url_faces = "https://raw.githubusercontent.com/Azure-Samples/cognitive-services-sample-data-files/master/ComputerVision/Images/faces.jpg"
# Select the visual feature(s) you want.
remote_image_features = ["faces"]
# Call the API with remote URL and features
detect_faces_results_remote = computervision_client.analyze_image(remote_image_url_faces, remote_image_features)

# Print the results with gender, age, and bounding box
print("Faces in the remote image: ")
if (len(detect_faces_results_remote.faces) == 0):
    print("No faces detected.")
    for face in detect_faces_results_remote.faces:
        print("'{}' of age {} at location {}, {}, {}, {}".format(face.gender, face.age, \
        face.face_rectangle.left, face.face_rectangle.top, \
        face.face_rectangle.left + face.face_rectangle.width, \
        face.face_rectangle.top + face.face_rectangle.height))

Detect adult, racy, or gory content

The following code prints the detected presence of adult content in the image. See Adult, racy, gory content for more details.

Detect Adult or Racy Content - remote
This example detects adult or racy content in a remote image, then prints the adult/racy score.
The score is ranged 0.0 - 1.0 with smaller numbers indicating negative results.
print("===== Detect Adult or Racy Content - remote =====")
# Select the visual feature(s) you want
remote_image_features = ["adult"]
# Call API with URL and features
detect_adult_results_remote = computervision_client.analyze_image(remote_image_url, remote_image_features)

# Print results with adult/racy score
print("Analyzing remote image for adult or racy content ... ")
print("Is adult content: {} with confidence {:.2f}".format(detect_adult_results_remote.adult.is_adult_content, detect_adult_results_remote.adult.adult_score * 100))
print("Has racy content: {} with confidence {:.2f}".format(detect_adult_results_remote.adult.is_racy_content, detect_adult_results_remote.adult.racy_score * 100))

Get image color scheme

The following code prints the detected color attributes in the image, like the dominant colors and accent color. See Color schemes for more details.

Detect Color - remote
This example detects the different aspects of its color scheme in a remote image.
print("===== Detect Color - remote =====")
# Select the feature(s) you want
remote_image_features = ["color"]
# Call API with URL and features
detect_color_results_remote = computervision_client.analyze_image(remote_image_url, remote_image_features)

# Print results of color scheme
print("Getting color scheme of the remote image: ")
print("Is black and white: {}".format(detect_color_results_remote.color.is_bw_img))
print("Accent color: {}".format(detect_color_results_remote.color.accent_color))
print("Dominant background color: {}".format(detect_color_results_remote.color.dominant_color_background))
print("Dominant foreground color: {}".format(detect_color_results_remote.color.dominant_color_foreground))
print("Dominant colors: {}".format(detect_color_results_remote.color.dominant_colors))

Get domain-specific content

Computer Vision can use specialized model to do further analysis on images. See Domain-specific content for more details.

The following code parses data about detected celebrities in the image.

Detect Domain-specific Content - remote
This example detects celebrites and landmarks in remote images.
print("===== Detect Domain-specific Content - remote =====")
# URL of one or more celebrities
remote_image_url_celebs = "https://raw.githubusercontent.com/Azure-Samples/cognitive-services-sample-data-files/master/ComputerVision/Images/faces.jpg"
# Call API with content type (celebrities) and URL
detect_domain_results_celebs_remote = computervision_client.analyze_image_by_domain("celebrities", remote_image_url_celebs)

# Print detection results with name
print("Celebrities in the remote image:")
if len(detect_domain_results_celebs_remote.result["celebrities"]) == 0:
    print("No celebrities detected.")
    for celeb in detect_domain_results_celebs_remote.result["celebrities"]:

The following code parses data about detected landmarks in the image.

# Call API with content type (landmarks) and URL
detect_domain_results_landmarks = computervision_client.analyze_image_by_domain("landmarks", remote_image_url)

print("Landmarks in the remote image:")
if len(detect_domain_results_landmarks.result["landmarks"]) == 0:
    print("No landmarks detected.")
    for landmark in detect_domain_results_landmarks.result["landmarks"]:

Get the image type

The following code prints information about the type of image—whether it is clip art or line drawing.

Detect Image Types - remote
This example detects an image's type (clip art/line drawing).
print("===== Detect Image Types - remote =====")
# Get URL of an image with a type
remote_image_url_type = "https://raw.githubusercontent.com/Azure-Samples/cognitive-services-sample-data-files/master/ComputerVision/Images/type-image.jpg"
# Select visual feature(s) you want
remote_image_features = VisualFeatureTypes.image_type
# Call API with URL and features
detect_type_results_remote = computervision_client.analyze_image(remote_image_url_type, remote_image_features)

# Prints type results with degree of accuracy
print("Type of remote image:")
if detect_type_results_remote.image_type.clip_art_type == 0:
    print("Image is not clip art.")
elif detect_type_results_remote.image_type.line_drawing_type == 1:
    print("Image is ambiguously clip art.")
elif detect_type_results_remote.image_type.line_drawing_type == 2:
    print("Image is normal clip art.")
    print("Image is good clip art.")

if detect_type_results_remote.image_type.line_drawing_type == 0:
    print("Image is not a line drawing.")
    print("Image is a line drawing")

Read printed and handwritten text

Computer Vision can read visible text in an image and convert it to a character stream. You do this in two parts.

Call the Read API

First, use the following code to call the batch_read_file method for the given image. This returns an operation ID and starts an asynchronous process to read the content of the image.

Batch Read File, recognize printed text - remote
This example will extract printed text in an image, then print results, line by line.
This API call can also recognize handwriting (not shown).
print("===== Batch Read File - remote =====")
# Get an image with printed text
remote_image_printed_text_url = "https://raw.githubusercontent.com/Azure-Samples/cognitive-services-sample-data-files/master/ComputerVision/Images/printed_text.jpg"

# Call API with URL and raw response (allows you to get the operation location)
recognize_printed_results = computervision_client.batch_read_file(remote_image_printed_text_url,  raw=True)

Get Read results

Next, get the operation ID returned from the batch_read_file call, and use it to query the service for operation results. The following code checks the operation at one-second intervals until the results are returned. It then prints the extracted text data to the console.

# Get the operation location (URL with an ID at the end) from the response
operation_location_remote = recognize_printed_results.headers["Operation-Location"]
# Grab the ID from the URL
operation_id = operation_location_remote.split("/")[-1]

# Call the "GET" API and wait for it to retrieve the results 
while True:
    get_printed_text_results = computervision_client.get_read_operation_result(operation_id)
    if get_printed_text_results.status not in ['NotStarted', 'Running']:

# Print the detected text, line by line
if get_printed_text_results.status == TextOperationStatusCodes.succeeded:
    for text_result in get_printed_text_results.recognition_results:
        for line in text_result.lines:

Run the application

Run the application with the python command on your quickstart file.

python quickstart-file.py

Clean up resources

If you want to clean up and remove a Cognitive Services subscription, you can delete the resource or resource group. Deleting the resource group also deletes any other resources associated with it.

Next steps

In this quickstart, you learned how to use the Computer Vision library for Python to do basis tasks. Next, explore the reference documentation to learn more about the library.