4.2.3. Azure AI Face Service
Azure AI Face (also called "Azure Face in Foundry Tools") specializes in facial detection and analysis. While Azure AI Vision handles general images, Face focuses specifically on human faces—finding them, analyzing their attributes, and (with restricted access) recognizing specific individuals.
Core capabilities:
| Capability | What It Does | Use Case |
|---|---|---|
| Face Detection | Finds faces in images | "Are there people in this photo?" |
| Face Attributes | Analyzes detected faces | "Is this person wearing glasses?" |
| Face Verification | Compares two faces | "Are these the same person?" |
| Face Identification | Matches face to database | "Who is this person?" (restricted) |
Face Detect API returns:
- Face rectangle: Bounding box coordinates for the face location
- Face ID: Unique temporary identifier for the detected face (24-hour expiration)
- Face landmarks: 27 precise positions of facial features (eyes, nose, mouth corners, eyebrows)
- Face attributes: Optional detailed analysis including:
- Age: Estimated age in years
- Emotion: Anger, contempt, disgust, fear, happiness, neutral, sadness, surprise
- Accessories: Glasses, headwear, mask detection
- Head pose: Roll, yaw, pitch angles (head orientation)
- Blur: Is the face blurry?
- Exposure: Is lighting appropriate?
- Noise: Is there visual noise?
- Occlusion: Is part of the face hidden?
Face landmarks explained: Face landmarks are precise coordinate points for facial features. These enable applications like:
- Virtual try-on (glasses, makeup)
- Facial tracking in video
- Emotion analysis from expressions
- Accessibility features for video calls
Example output structure:
{
"faceId": "abc123-temp-id",
"faceRectangle": {"top": 100, "left": 50, "width": 80, "height": 100},
"faceAttributes": {
"age": 32,
"emotion": {"happiness": 0.95, "neutral": 0.05},
"accessories": [{"type": "glasses", "confidence": 0.98}]
}
}
Access restrictions: Due to privacy and ethical concerns, Microsoft restricts access to certain Face features:
- Unrestricted: Basic face detection, face rectangle, attributes
- Restricted: Face verification, face identification, face grouping
- Application required: Recognition features require applying for access and demonstrating legitimate use
⚠️ Exam Tip: Questions about "accessories," "glasses," "headwear," or "head pose" point to Face Attributes. Questions about "is this the same person" point to Face Verification. Questions about "who is this person" point to Face Identification (which is restricted).
⚠️ Important: Face recognition features (identifying WHO a person is) have restricted access due to privacy concerns. Basic detection and attribute analysis are more broadly available.
