Overview

The SDKs use AI to automate the processing of large numbers of images. This improves the efficiency of image processing, including portrait production and retouching, commodity image editing and publishing, and marketing poster production. You only need to call the API and upload images. The SDKs are developed based on the best practices of Alibaba Group in image segmentation. They process billions of SKUs on an annual basis and can automatically produce commodity images with white backgrounds. This saves the Alibaba platforms and its customers more than ten million hours and tens of millions of USD on overhead costs annually.

• Fast Implementation Download a demo APK to test the SDKs before you make a purchase. Simply integrate the SDKs with your applications after official purchase.
• Data Security All data is locally processed on devices to maximize data security.
• Open Ports The API can meet your business requirements in most scenarios.
• Automation After subjects and images are detected or recognized, they are automatically classified to trigger real-time actions. This improves user experience.
• Precise Recognition High-precision models and algorithms help recognize images and detect subjects.
• Intelligent Optimization Intelligently detect and identify a variety of objects such as object and backgrounds, and automatically generate tags and classify the subjects.

Features

Real-time Video Segmentation

The SDK enables high-precision image segmentation based on a deep learning framework and detection and recognition technologies. The SDK can enable automatic subject recognition and pixel-level scene recognition in real time. The SDK is suitable for scenarios in which subjects in images consist of hollow objects or images contain a complex background.

Real-time Lossless Segmentation

The SDK splits live videos in real time based on color filtering and deep learning technologies. The SDK also smoothens the edges of video images and removes noises from video images. This way, live videos can be split into segments whose quality meets the requirements of enterprises in the film and television industry.

Offline Image Segmentation

The SDK performs fine-grained and flawless image keying on pictures that you capture or upload based on detection and recognition technologies. The SDK can help you enable high-precision image segmentation and cut and replace elements in images with complex backgrounds.

Offline Video Segmentation

The SDK splits collected video images in a serialized manner by using on-device computing resources. The offline image enhancement feature ensures that image quality does not decrease when a user zooms in on an image to a size that is twice the original size of the image.

Filter

The following filters are provided: normal, energetic, fresh, food, Japanese style, retouching, mint, and black and white. The SDK edits images to apply the specified effects without impact to the quality of the images.

Retouching

The SDK performs retouching operations such as skin smoothing, sharpening, and whitening, and precisely edits facial features based on fine-grained facial landmark recognition technologies. The SDK is suitable for applications such as entertainment live streaming applications and beauty camera applications.

Offline Image Enhancement

The SDK uses the super-resolution algorithm developed by Alibaba Cloud to enhance image details in real time on mobile clients, remove various defects such as noise and burrs, optimize image parameters such as exposure, color temperature, and contrast, and significantly improve the subject perception of images.

Offline Gesture Recognition

The SDK tracks and recognizes gestures in real time, and displays hand information in real time, including the type of gesture, coordinates of the gesture box, palm orientation, judgment of left and right hands, number of finger landmarks, and 2D and 3D coordinates of finger landmarks.

Offline Body Landmarks

The SDK provides 15 body landmarks that are customized based on the characteristics of human bodies and can accurately depict the postures of characters. The SDK ensures clarity against low ambient light and blurs.

Offline Body Movement Counting and Feedback

The SDK captures human body movements by using cameras, recognizes body landmarks in real time, and automatically counts the movements. The SDK supports 14 types of fitness movements and allows users to customize fitness movements. In addition, the SDK can detect 15 types of irregular movements and notify users of the irregular movements in real time.