ShiMeta AI Recognition Algorithms
Product Overview
ShiMeta’s AI algorithm platform integrates high-accuracy face, video, and object recognition algorithms, providing embedded smart terminal solutions for security, retail, industrial monitoring, and intelligent advertising.
Built for Android, Linux, and embedded systems, these algorithms are edge-optimized and support offline operation, enabling fast, secure, and responsive decision-making without relying on the cloud.
Face Recognition-Commercial Display
Algorithms include face detection, face attributes, face search, monocular vivo, binocular vivo, etc.
Face Recognition-Personal ID
Algorithms include face detection, evidence comparison, face search, monocular live, binocular live, etc.
Key Functions:
- Face Detection & Tracking
Detect and track faces in static images or video in real-time using visible light or near-infrared cameras.
- 1:1 and 1:N Face Matching
Local face library for real-time face comparison, supporting high-speed authentication.
- Liveness Detection
Anti-spoofing protection with RGB (monocular), RGB+IR (binocular) liveness detection.
- ID Match Verification
Compare real-time face with stored ID photo to validate user identity.
- Attribute Analysis
Recognize gender, age, mask status, glasses, helmet presence—ideal for public-facing terminals.
Applications:
- Smart advertising display
- Face access control systems
- Facial attendance machines
- Customer traffic analytics
Video Behavior Recognition Algorithm Suite
Functional Coverage (30+ Models Supported):
Category | Sample Capabilities |
Behavioral Monitoring | Off-duty detection, distress gestures, falling, unusual loitering, climbing |
Crowd Control | Crowd density, reverse walking, illegal intrusion |
Safety Monitoring | Helmet, uniform, smoke, fire, phone, smoking |
Vehicle Monitoring | License plate recognition, parking violation, reversing |
Smart Alerts | Real-time camera analysis, multi-zone alarm triggering, automatic recording |
Motion Recognition
Video analysis is performed on the images captured by the camera, and predefined combinations of actions are made by continuously monitoring the human body behavior in the images, and the recognition of (continuous) actions is judged according to pre-defined rules. The algorithm is customized according to the requirements.
Fall Detection
Real-time analysis of the images captured by the camera, can set the area preset range of interest, deployment time, and generate an alarm when the person entering the set area produces a fall behavior.
Climbing Detection
Through the camera monitoring, the images captured by the camera are analyzed in real time, and the area can be set to preset the scope of interest, the deployment time, and the alarm will be generated when the person entering the set area generates climbing behaviors (over the fence, railing, etc.).
SOS Testing
Through the camera monitoring, the images captured by the camera are analyzed in real time, and the area can be set to preset the range of interest, the deployment time, and the alarm is generated when the person entering the set area produces a distress behavior (raising his hands above his head and waving his hands).
Duty departure recognition
Real-time analysis of the images captured by the camera, regional preset scope of interest, deployment time, intelligent identification of whether the people in the image of the departure behavior, when found that the staff is not in the work area for a period of time exceeding the set time threshold, then the alarm.
Anomalous Wandering Recognition
Real-time analysis of the images captured by the camera can be set to preset the area of interest, deployment time, length of stay, real-time tracking of people entering the set area, and generate an alarm when the length of stay exceeds the set threshold.
Crowd identification
Real-time analysis of the screen captured by the camera, can set the area preset range of interest, deployment time, intelligent identification of the screen in the set area whether the number of people in the set threshold, more than the number of people will be alarmed.
Pedestrian against traffic
Through the camera monitoring, the screen captured by the camera is analyzed in real time, and the area can be set to preset the scope of interest, the deployment time, and the direction of travel of the personnel entering the set area is analyzed, and an alarm is generated if the direction of travel is inconsistent with the stipulated direction of travel.
Trespassing Recognition
Preset the range of interest and arming time for the captured image area. Parses the video stream in real time and generates an alarm event when someone enters the area within the set deployment time.
Headcount
The location and number of pedestrians' heads are analyzed in real time from the images captured by the camera.
license plate recognition
Algorithm contains vehicle detection, license plate number recognition, through the camera to obtain the image, first detect the vehicle location, and then identify the license plate number in the vehicle area; support the main types of license plate numbers in the mainland.
Vehicle inspection
Getting pictures from cameras, detecting vehicles in real time and locating and acquiring vehicle areas in real time.
Workwear Testing
Recognizes whether the person in the screen is wearing uniforms or not. Generates an alarm and an event when a person entering the set area generates behavior without uniforms, and uniforms can be entered by passing in a photo of the person wearing uniforms.
Helmet Testing
Analyzes whether the operator in the screen is wearing a helmet correctly, generates an alarm and generates an event when a person entering the set area produces no helmet.
Call Testing
Analyzes the screen to see if someone is making a phone call, generates an alarm and generates an event when a person entering the set area generates a phone call.
Smoking Detection
Analyzes the screen to see if someone is engaged in smoking behavior, generates an alarm and generates an event when a person entering the set area generates a smoke.
pyrotechnic detection
Recognize whether there is smoke, fire and other abnormalities in the screen. In the presence of smoke and fire into the set area, generate an alarm and generate events, you can capture the scene picture custom development to improve the recognition accuracy.
Features:
- Simultaneous multi-model deployment
- Customizable area-of-interest and trigger conditions
- High-precision alarm accuracy with minimal false positives
Scenarios:
- Factories, mines, and construction zones
- Campuses and smart cities
- Warehouses and secure areas
Object Recognition Algorithm Suite
Core Capabilities:
- Product Image Recognition
Recognize vegetables, snacks, fruit, bottled items instantly from camera feeds.
- Barcode & QR Code Reading
Auto-detect and extract all embedded text and data from visual codes.
- Text Character Recognition (OCR)
Real-time scanning and output of printed or digital text within images.
Features:
- 9% recognition accuracy
- Sub-0.1 second identification time
- Pre-trained product libraries with offline support
- Clean and simple Android-compatible API integration
Applications:
- Smart retail weighing and checkout
- Vending machines
- Shelf-label monitoring systems
Commodity Identification
Through the images collected by the camera, analyze the images of bottled goods in the screen in real time, recognize the types of bottled goods in the images, and give the recognition results, supporting vegetables, fruits, dry goods, snacks and so on.
Barcode recognition
Automatically detects barcodes in images and analyzes and recognizes the text information contained in the barcodes.
QR code recognition
It is possible to recognize whether or not the image contains QR code information, and output text information (characters such as URLs or text corresponding to each QR code) contained in the QR code in the image.
text recognition
Real-time analysis of the position of the text box on the screen and the content of the text.
Developer Tools & Integration
- Android & Linux system compatibility
- Support for C/C++ and Java SDKs
- Lightweight APIs for rapid deployment
- Documentation and sample code included
Software Delivery & Deployment
- Edge Device Integration: All models designed for terminal-side execution
- Cloud-Free Operation: Works offline or within closed networks
- Custom Model Support: Scalable to your specific project requirements
Ideal for:
- Smart Terminal Integrators
- Security System Providers
- Smart Retail Technology Firms
- AIoT Platform Developers
- Surveillance System Manufacturers