Overview of Face and Body features, activation, and scenarios-Visual Intelligence API(VIAPI)-阿里云帮助中心

The Alibaba Cloud Visual Intelligence API uses facial and body analysis technology developed by DAMO Academy to provide features such as face detection and landmark localization, facial attribute recognition, face comparison, face search, body detection, body attributes, and behavior analysis. It offers high-performance and high-availability (HA) facial and body recognition services for developers and enterprise users. The service is widely used in fields such as digital outlets, building access control, identity verification, interactive entertainment, IP cameras, and content advertising.

Service activation

You can click Activate Now on the product page to activate the service. For more information, see Quick start.

Features

Category	Capabilities	Description
Face detection	Face detection and landmark localization	Detects faces in an image and returns the coordinates of the bounding box for each detected face.
Facial attribute	Facial attribute recognition	Based on face detection, this feature recognizes four attributes for each detected face: gender (male or female), age, expression (smiling or not smiling), and glasses (wearing or not wearing).
Facial attribute	Expression recognition	Detects and recognizes facial expressions in an image. The supported expressions are neutral, happiness, surprise, sadness, anger, disgust, and fear.
Face editing	Face enhancement	Detects faces in an image, then crops, snaps, and enhances the details of the faces. The enhanced faces are then merged back into the original image.
	Face retouching	Retouches faces in an image. This includes skin smoothing, whitening, and removing dark circles and nasolabial folds.
	Face shaping	Shaping adjusts the contours and features of a face.
	Face anonymization	Blurs faces in an image and outputs the anonymized image.
	Anime-style character generation	Transforms a character image into a 2D cartoon style and returns the cartoonized image.
	Portrait sketch stylization	Automatically crops the head region of a character image and generates a sketch effect for the headshot.
	Smart face slimming	Analyzes the facial features in a character image and generates an image with adjusted facial features and a slimmer face. It can process up to three faces in an image with multiple people.
	Smart skin beautification	Automatically evens skin tone, removes blemishes such as acne, acne scars, and freckles, and whitens the skin of the entire body in a character image. It preserves skin texture as much as possible. This feature supports images with multiple people.
	Face fusion	You can optimize areas of a person's body in an image, such as the neck, torso, and limbs, to render a more aesthetically pleasing body shape that appears taller, slimmer, and more refined, while preserving smooth body lines, natural curves, and balanced proportions. This feature supports processing for multiple people and various body shots, such as full-body and half-body.
	Add a face fusion template	Adds a face image that has passed content moderation as a template for the face fusion feature.
	Query face fusion templates	Queries created face templates.
	Delete a face fusion template	Deletes created face templates.
Body recognition	Hand gesture keypoint detection	Obtains the 21 keypoints of a hand gesture.
	Body pose keypoint detection	Obtains the 18 keypoints of a human body.
	Body detection	Detects human bodies in an image.
	Action and behavior recognition	Recognizes human actions and behaviors in videos and images and returns the recognized behavior category. Currently supported categories include raising a hand, eating or drinking, smoking, making a phone call, using a mobile phone, sleeping at a desk, and falling.
	Body counting	Counts the faces in the input image.
	Structured body attributes	Detects the attributes of human bodies in an image. This includes body detection and attribute estimation, such as gender, age, orientation, hat, glasses, bag, clothes, and color.
	Pedestrian intrusion detection	Uses a deep learning algorithm to detect if a person has entered a monitored perimeter or area in a video.
	Static hand gesture recognition	Recognizes hand gestures in an image.
	Online proctoring	Recognizes the behavior of candidates during an online exam. It supports the detection of chat tools on the screen and the candidate's status.
	Fingerprint extraction	Automatically locates the finger area in a photo of a finger and generates a binarized, black-and-white, press-type fingerprint image.
Liveness detection	Face liveness detection	Detects whether the living object (mainly a face) in an image is from a direct capture or a recaptured photo. A face must be present in the image for liveness detection to work.
	Video liveness detection	Detects whether the face in a video is from a direct capture or a recaptured video.
	Infrared face liveness detection	Detects whether the face in an infrared image is a live human face captured at close range by an authenticated device. It can be widely used in real-time face capture scenarios with infrared devices to meet the authenticity and security requirements of infrared face registration and authentication. An infrared image must contain a face for this feature to work.
Facial recognition	Public figure recognition	Recognizes public figures in an image.
	Celebrity recognition	Recognizes celebrities in an image.
	1:1 face comparison	Based on two input images, this feature detects the faces in both images and compares the largest face from each to determine if they belong to the same person.
	Create a face database	Creates a face database.
	View database list	Views the list of face databases.
	Add a face entity	Adds face entity data to a face database.
	Query a face entity	Queries face entity data in a face database.
	Query a list of face entities	Queries the list of face entities in a face database.
	Update a face entity	Updates face entity data in a face database.
	Add face data	Adds face data to a specified database.
	Face search	Searches a database for similar face images based on an input image.
	Delete a database	Deletes a specified face database.
	Delete a face	Deletes face image information from a specified database.
	Delete a face entity	Deletes face entity data from a face database.
	1:1 face comparison with masks	Based on two input images, this feature selects the largest face from each image and compares them to determine if they belong to the same person. It integrates three optimized technologies: masked face generation, occlusion-resistant keypoint localization, and occlusion-resistant feature attention, to achieve high-speed facial recognition even when a person is wearing a mask.
Financial-grade face detection	Server-side identity verification	Requests user image information for identity verification on the server-side based on the user's name and ID number.
	Mobile client request for identity verification	Requests user image information for identity verification on a mobile client based on the user's name and ID number.
	Mobile client query for identity verification	After confirming that the certificate is valid and the name matches, this feature returns the user's image information for identity verification.

Scenarios

The following are common scenarios for the Face and Body service:

Identity verification
The service performs detection based on image or video input. It compares faces against a registration database for 1:N facial recognition or against a certificate for 1:1 face-to-certificate comparison. This is suitable for scenarios such as face logon, VIP facial recognition, and facial recognition-based access control.
Security monitoring
The service uses facial landmark localization and tracking to perform 3D face reconstruction and estimation. It recognizes actions such as shaking the head, blinking, and opening the mouth to determine if the person is real. This is used for secure identity verification in scenarios such as security checks and applications.

For more product updates, follow the Alibaba Cloud Visual Intelligence API.