Image segmentation
Built on Alibaba Cloud's deep learning technology, image segmentation integrates detection and recognition technologies to deliver high-precision visual segmentation.
The image segmentation service performs fully automatic, pixel-level recognition of subjects and scenes within seconds to create four-channel transparent material. It delivers excellent results for fine-grained details, such as individual strands of hair, subjects with hollow areas, and complex backgrounds. The service supports segmentation for people, products, and general scenes, and is widely used in e-commerce, retail, entertainment, and personal applications.
Activate the service
Click Activate Now on the product page to activate this service. For detailed instructions, see Quick Starts.
Features
The Alibaba Cloud Vision AI Platform provides the following image segmentation features:
Category | Feature | Description |
portrait segmentation | Identifies the contours of human bodies in an input image, separates them from the background, and returns a four-channel image of the foreground portrait. This feature is suitable for single or multiple people, complex backgrounds, and various body poses. | |
Identifies facial contours in an input image, excluding the neck, ears, and hair, and returns a four-channel transparent image that shows only the face region. This feature is suitable for single-person and multi-person scenarios. | ||
Identifies hair outlines in an input image, excluding the neck and ears, and returns a four-channel transparent image that shows only the hair region. This feature is suitable for single-person and multi-person scenarios. | ||
Identifies and segments the skin areas of people in an image. | ||
Commodity segmentation | Identifies product outlines in an input image, separates them from the background, and returns a four-channel image of the foreground products. This feature is suitable for single or multiple products and complex backgrounds. | |
Performs pixel-level segmentation on clothing in an input image and returns the segmentation result. | ||
General segmentation | Automatically identifies the main subject in an input image, separates it from the background, and returns a four-channel image of the foreground subject. This feature is suitable for segmenting people, animals, food, and other objects. | |
Identifies the sky region in an input image, separates it from the background, and returns the segmented foreground region. | ||
Performs pixel-level segmentation on food items in an input image and returns the segmentation result. | ||
Refines a coarse mask for an input image and outputs a high-precision mask. | ||
Performs pixel-level segmentation on the sky in an input image. | ||
Segments the main subject in an image and outputs a corresponding transparent PNG file. | ||
Identifies the contours of human bodies in an input image, separates them from the background, and returns the segmented foreground portrait. | ||
segmentation and replacement | Replaces the sky in a target image with the sky style from a reference image. |
Use cases
Common use cases for image segmentation include:
E-commerce banner creation
Combine segmentation with templates to quickly generate e-commerce marketing images in various styles. This approach significantly reduces design costs and improves efficiency for merchants.
ID photo creation
Use portrait segmentation to separate a person from the background. Then, combine it with the image super-resolution feature from the Alibaba Cloud Vision AI Platform to enhance image quality and create professional ID photos.
For more product updates, follow the Alibaba Cloud Visual Intelligence API.