​

Computer Vision (Image Analysis)

​

Key Capabilities

​

Image Analysis Features

​

Read Text from Images (OCR)

• Extract printed and handwritten text
• Support for multiple languages
• High accuracy text recognition
• Faster performance than async Read API
• Returns text with bounding box coordinates

from azure.ai.vision.imageanalysis import ImageAnalysisClient
from azure.core.credentials import AzureKeyCredential

client = ImageAnalysisClient(
    endpoint="https://<your-resource>.cognitiveservices.azure.com/",
    credential=AzureKeyCredential("<your-key>")
)

result = client.analyze_from_url(
    image_url="https://example.com/image.jpg",
    visual_features=["READ"]
)

for block in result.read.blocks:
    for line in block.lines:
        print(f"Text: {line.text}")

​

Generate Image Captions

• Simple captions: One-sentence descriptions of the entire image
• Dense captions: Detailed captions for individual objects with bounding boxes
• Natural language descriptions
• High accuracy based on image content

result = client.analyze_from_url(
    image_url="https://example.com/image.jpg",
    visual_features=["CAPTION"]
)

print(f"Caption: {result.caption.text}")
print(f"Confidence: {result.caption.confidence}")

​

Detect Objects

• Detect multiple instances of the same object
• Return pixel coordinates for each object
• Confidence scores for each detection
• Support for thousands of object categories

result = client.analyze_from_url(
    image_url="https://example.com/image.jpg",
    visual_features=["OBJECTS"]
)

for obj in result.objects:
    print(f"Object: {obj.tags[0].name}")
    print(f"Confidence: {obj.tags[0].confidence}")
    print(f"Bounding box: {obj.bounding_box}")

​

People Detection

• Returns bounding box coordinates for each person
• Confidence scores for each detection
• Works with single or multiple people

result = client.analyze_from_url(
    image_url="https://example.com/image.jpg",
    visual_features=["PEOPLE"]
)

for person in result.people:
    print(f"Person detected at: {person.bounding_box}")
    print(f"Confidence: {person.confidence}")

​

Tag Visual Features

• Thousands of recognizable objects, living things, scenery, and actions
• Tags with confidence scores
• Context hints for ambiguous tags
• Includes both main subjects and background elements

result = client.analyze_from_url(
    image_url="https://example.com/image.jpg",
    visual_features=["TAGS"]
)

for tag in result.tags:
    print(f"{tag.name}: {tag.confidence}")

​

Smart Cropping (Area of Interest)

• Analyzes image content to determine focus area
• Returns bounding box coordinates
• Supports custom aspect ratios
• Preserves the most important visual elements

result = client.analyze_from_url(
    image_url="https://example.com/image.jpg",
    visual_features=["SMART_CROPS"],
    smart_crops_aspect_ratios=[0.9, 1.33]
)

for crop in result.smart_crops:
    print(f"Crop box: {crop.bounding_box}")

​

Additional Features (v3.2)

• Brand Detection: Identify commercial brands and logos
• Face Detection: Detect faces and estimate age and gender
• Image Type Detection: Determine if image is clip art or line drawing
• Color Scheme Detection: Identify dominant and accent colors
• Adult Content Detection: Detect adult, racy, or gory content
• Domain-Specific Models: Detect celebrities and landmarks
• Image Categorization: Categorize images using a taxonomy

​

Multimodal Embeddings

• Vectorize images for similarity search
• Convert text queries to vectors
• Match images to text based on semantic meaning
• Support for 102 languages (multilingual model)
• Build image search applications

from azure.ai.vision.imageanalysis import VectorizeClient

vectorize_client = VectorizeClient(
    endpoint="https://<your-resource>.cognitiveservices.azure.com/",
    credential=AzureKeyCredential("<your-key>")
)

# Vectorize an image
image_vector = vectorize_client.vectorize_image(
    image_url="https://example.com/image.jpg"
)

# Vectorize a text query
text_vector = vectorize_client.vectorize_text(
    text="a dog playing in the park"
)

​

API Versions

​

Version 4.0 (Recommended)

• Synchronous OCR (Read)
• People detection
• Dense captions
• Enhanced image captioning
• Improved smart cropping
• Multimodal embeddings

​

Version 3.2

• Async OCR (Read API)
• Brand detection
• Face detection
• Celebrity and landmark detection
• All other v3.2 features

​

Input Requirements

• Supported formats: JPEG, PNG, GIF, BMP, WEBP, ICO, TIFF, MPO
• File size: Less than 20 MB
• Dimensions: 50 x 50 to 16,000 x 16,000 pixels

• Supported formats: JPEG, PNG, GIF, BMP
• File size: Less than 4 MB
• Dimensions: 50 x 50 to 16,000 x 16,000 pixels

​

Region Availability

​

Use Cases

​

Getting Started

pip install azure-ai-vision-imageanalysis

​

SDK Support

pip install azure-ai-vision-imageanalysis

dotnet add package Azure.AI.Vision.ImageAnalysis

<dependency>
  <groupId>com.azure</groupId>
  <artifactId>azure-ai-vision-imageanalysis</artifactId>
</dependency>

npm install @azure-rest/ai-vision-image-analysis

​

Pricing

• Free Tier (F0): 5,000 transactions per month
• Standard Tier (S1): Pay per 1,000 transactions
• Different pricing for v3.2 and v4.0 features
• Additional costs for custom models

​

Next Steps

• Try Vision Studio
• View Image Analysis Quickstart
• Explore Code Samples
• Learn about Custom Vision