Aggregate artificial intelligence technology to power the deep integration of native technology and the industry
Aggregate artificial intelligence technology to power the deep integration of native technology and the industry
Support the fine-grained recognition of 14 types and 20000 models of vehicles with the accuracy greater than 95%.
Support the accurate recognition of 14 colors of car bodies, 10 types of vehicle markers such as registration tag, decoration and sun shield, 3 kinds of dangerous driving behaviors including talking on the phone, failing to fasten safety belt and smoking, vehicle types recognition for toll collection and tonnage recognition of heavy goods vehicles.
Based on deep learning technology, our product supports the recognition of 6 license plate colors, 14 kinds of license plate types and different provincial license plates.
Extracting the individual characteristics of each vehicle’s appearance, our product can realize the recognition of fake license plate car and unlicensed car, so that unlicensed car can also access the unmanned parking lot service and toll collection service.
Support locating 25, 68, 106 high-precision facial keypoints which clearly mark the brow, eyes, mouth, nose and face outline.
Analyze 5 face related attributes including gender, age, race, beauty score and smile to utilize crowd data effectively.
Determine the identity of given face by searching in database (Provide FAR and VR). Adapt to various face postures, lighting scenes and overlapped/blurred scenes.
Detect faces in photos or videos and return high-precision face bounding boxes. Tracking its route and mark down its trajectories.
Determine if the person in front of the camera is alive. Point out photos and models from alive people to prevent identity fraud and ensure the liveness.
Convenient and efficient identity proofing with face photo verification reduces security risks effectively.
Locate the landmarks of human body such as head, trunk, limbs and hands. Support human motion recognition and pose recognition of multiple people, multiple views and variant poses.
Pixel-level body attributes and accessories recognition. Accurately analyze human characteristics such as gender, age, and body shape, and finely recognize external features such as clothing and accessories to accurately describe multiple attributes of people.
Detect and track moving targets in real time. Draw specific motion trajectories. Generate structured target information. Analyze movement routes and behaviors.
Our product can not only detect the large defects such as holes and oil stains which need to be cut, but also detect minor defects such as crease mark, double yarn, missing yarn, broken yarn and spot. It is convenient for workers to evaluate the grade of textile and provide the basis for the next process.
Spinning cake monitoring equipment can quickly and efficiently detect defects such as wool, stiff silk, tripping silk, tailless and oil stains, and record them in the spinning cake files, which is convenient for evaluating the grade of spinning cake and providing a basis for checking the output and quality of spinning cake in the future.
The food recognition algorithm is based on deep learning. It can recognize about 1,000 kinds of foods and accurately locate them. The recognition accuracy is more than 90%. Besides, it can propose personalized healthy diet suggestions for users.
Accurately recognize the location, brand, name and price of shelf goods, so as to facilitate timely and accurate grasp of shelf goods information and improve the efficiency of supermarket shelf management.
Accurately locate and recognize the household items to realize the functions of smart advertising based on the content of the scene.
Accurately analyze and describe the breed, name, hair characteristics and some other characteristics of pets to help locate and recognize pets in multiple scenes.
Automatically recognize and sort solid waste to realize the material recognition and find the optimal position to pick solid waste up.
Real-time head-shoulder detection and tracking system is based on deep learning. It supports the accurate head-shoulder detection and tracking in HD video and draws trajectories for each specific target to provide real and efficient underlying data for further big data analysis.
Real-time object detection and tracking system is based on deep learning. It aims at detecting and tracking specific targets in the real-time HD surveillance video with complex background.
Support real-time target detection of pedestrian, bicycle, electric vehicle, motorcycle, car and other 11 types of targets by multi-channel HD video.
Track all targets in real time. Draw specific trajectories for each target, and generate structured target information.
Automatically recognize the text contained in images for arbitrary scenes. The extracted text content can provide highly relevant semantic information for users to make the management more convenient and efficient. It has a wide range of applications in human-computer interaction and VR fields.
Identify the shelf identification number. To achieve effective management, grasp the container information and arrangement information of each shelf and carry out zoning and hierarchical management.
Recognize the content of various types of bills to realize the rapid input of a large number of paper bills so as to improve efficiency and reduce cost.
The automatic recognition of meter readings can be used in various complex environments and return recognition confidence.
Automatically extract the text information on the front and back of the certificate. Recognize the number of the certificate and the issuing authority. Then provide authenticity judgment of the certificate.
