Magic Cube Smart Analysis Box

XCore Magic Cube Smart Analysis Box Ecological algorithms


Video Structuring

Face and Keypoint Detection

Quickly detects faces and marks face coordinates, extracts key points for face alignment including cheeks, eyebrows, eyes, mouth, nose and other five facial features, accurately identifies multiple attribute information, and provides reliable support for face recognition. It can run on multiple platforms at the end and side to achieve efficient, accurate and stable face detection function.

Face and Keypoint Detection

Based on the ai algorithm framework, it supports extraction and analysis of face features in various complex scenes (indoor, outdoor, strong light, low light) to accurately complete face recognition, and reduce the extraction difficulty and release greater computational performance through multiple video stream inference; application scenarios cover retail payment, intelligent security, access control, time and attendance, identity verification, etc.

Face attribute analysis

According to the AI algorithm to calculate the face features and face-related attribute analysis, including gender, age, emotion, etc.

Pedestrian attribute analysis

AI algorithm pedestrian detection to obtain high precision recognition of pedestrian attribute information and action, including gender and age, clothing category, clothing color, wearing object, behavior and action, etc., providing rich structured information, which can be used for efficient video material management and accurate marketing.

Pedestrian attribute analysis
License plate/model detection and recognition

Through AI algorithm, detect the location of the subject vehicle, license plate (license plate frame 4 vertices) / vehicle brand model recognition, while supporting multiple video stream inference, suitable for park, traffic, parking lot and other scenarios.

License platemodel detection and recognition
Helmet and other object detection

Detection of helmets and faces according to pedestrian attribute analysis algorithm, through video structuring to determine whether to wear helmets, support including red, white, yellow, blue, orange helmet recognition training; at the same time the algorithm can be used for the detection and recognition of a variety of objects, the accuracy rate is higher than 95%, effectively improve the safety supervision and management work.

Helmet and other object detection​
Pet identification

AI algorithm can recognize a variety of pet species and support multi-species recognition of a single species; application scenarios are applicable to pet machines, photo-recognition, early childhood science, image content analysis and other scenarios.

Pet identification​
Elevator electric car recognition warning

Detection and identification of all types of electric vehicles, when the detection of electric vehicles into the elevator, the camera can be directly alarm (sound / flash), and automatically upload the alarm information, live images, video and other data to the cloud platform, the management personnel can receive timely information, view the scene, for timely processing.

Work uniform recognition

Detection and identification of various types of uniforms, in the office or construction site, the camera can directly monitor whether the personnel in it wears the specified clothing, and detects the personnel who are not wearing the specified uniforms and automatically uploads the alarm information, site images and to the cloud platform, the management personnel can receive information, view the site situation and carry out timely processing.

Work uniform recognition​
Throwing objects from height

AI algorithm can monitor the aerial throwing event 24 hours a day, can effectively filter trees / flying birds / clouds / wind / rain and snow weather / vibration and other sources of interference, the algorithm accuracy is greater than 90%.

Throwing objects from height​
Mask / hood recognition

Enter and exit the area according to the provisions of hoods, masks, work shoes, such as non-compliance, the system automatically take pictures and form evidence files, algorithm accuracy greater than 90%.

face mask detection
Non-motorized vehicle capture

By identifying the non-motorized vehicles entering the area, it can display the capture picture and capture time when a non-motorized vehicle is detected.

Chef's hat/chef's uniform recognition

By identifying the chefs entering the back kitchen space, when a chef’s hat/chef’s uniform is not detected, it can immediately alert and generate corresponding pictures and video clips.

Chef's hat/chef's uniform recognition​
Road damage recognition

Real-time recognition of road surface damage, when detecting deep potholes, collapses, cracks or road depressions in the road, immediately alert and generate corresponding pictures and video clips.

Road damage recognition​
Vehicle No Parking

Real-time recognition of road no-parking sections and generation of corresponding pictures when illegal parking is detected

Warning Line

Real-time identification of prohibited road sections, when the detection of people crossing the border, alarm capture pictures

Person intrusion

Real-time identification of each security rights, when found someone illegal entry, alarm capture pictures

Person wandering

Real-time identification of people walking status, when found in the lingering wandering, this person key mark, generate pictures

Over the wall

Real-time recognition of wall status, when found someone illegally over the wall, alarm generation picture


Tracking statistics

Pedestrian detection

Efficient detection and tracking based on pedestrian coordinates, supporting multiple video streams with simultaneous inference, realizing target tracking for high frame rate, low frame rate and ultra-low frame rate, adaptive adjustment of tracking parameters, balancing processing speed and algorithm accuracy, providing support for pedestrian attribute analysis and pedestrian re-identification.

Pedestrian detection​
Pedestrian re-identification

Based on AI algorithm pedestrian detection to obtain its features, structured storage; when need to find, the system will extract the features of the target person, match with the image features already stored in the database and identify the identity of the target task.

Pedestrian re-identification​
Crowd density analysis

Using AI algorithm of face recognition and pedestrian re-identification technology to turn “people” into “data”, support real-time display of crowd trends in different time periods and crowd gathering area portrait video stream inference; can be applied to scenarios: railway stations, office halls, pedestrian streets etc.

Crowd density analysis​
Passenger flow statistics

Passenger flow statistics is combined with face/head detection AI algorithm to help stores establish a portrait of customers visiting the store from dimensions such as gender, age, expression, new and old customers, and length of stay; deeply digging into user needs to serve as a basis for adjusting store layout.

Crowd density analysis

Behavioral posture

Fall recognition

Based on the behavior posture recognition technology, it automatically identifies the fall behavior of people in public places, and supports fall recognition in multiple scenes and different backgrounds, which is applicable to the fall warning of the elderly and children in public places.

fall recognition
Perimeter detection

Based on AI algorithm human body recognition technology, accurate human body detection analysis identification for personnel, to achieve early warning of dangerous areas / protection areas of personnel intrusion detection; can customize the perimeter drawing, support multi-person scenario early warning.

Perimeter detection
Employee departure detection

Customizable area range to detect the head in the range, and count the head in the perimeter, according to whether there is a qualified head in the perimeter to determine whether the employee is off duty.

Playing with cell phones

Based on behavioral posture recognition technology, it automatically identifies cell phone playing behavior and supports multiple scenes and different backgrounds of cell phone playing, which is suitable for early warning of office areas and meetings.

Playing with cell phones

Based on behavior posture recognition technology and smoke alarm behavior, automatic recognition of smoking recognition, and support for multiple scenes, smoking in different contexts, suitable for smoke-free, non-smoking environment alarm and early warning


Based on behavioral posture recognition technology, it automatically identifies sleeping posture and supports sleeping in multiple scenes and different contexts, which is suitable for alarm as well as early warning during work

Answer the phone

Based on behavioral posture recognition technology, automatic identification is during work, and support multiple scenes, sleeping in different contexts, suitable for alarm during work as well as early warning

listen mobile

Human body recognition technology based on AI algorithms for accurate action analysis of people and early warning detection of brawls in dangerous areas/protected areas;


Human body recognition technology based on AI algorithms for accurate action analysis of people for capture;

xcore smart ai box
xcore smart ai bos

Product Details


If you want to know more about the algorithm, or if you have your own demand scenario, fill out the form below to get a customized solution!

Packaging customization

    Application Scenarios XCORE

    Smart Product Solutions