Products Description
1. Product Overview
The AI Edge Computing Server Box (AIbox-05) is a smart application server integrating over 100 AI algorithms. Featuring a quad-core Cortex-A55 processor with a maximum clock speed of 2.2GHz, Mali-G52 MC3 GPU, and 6T computing power, it enables rapid deployment of AI capabilities at the edge. The plug-and-play design delivers instant AI empowerment.
2. Product Features
● Uses a variety of AI-specific algorithms.
Supports RTSP video streaming and GB28281 protocol access.
● Enables the integration of computing power and application middleware for unified edge deployment.
●Support remote device upgrade (device connected to the network)
●Open interface, supports secondary development and docking
Supports AI algorithm customization.
●Supports multi-channel video stream convergence management (NVR)
●Supports assigning different algorithm analysis tasks to different cameras
3. Specifications and Parameters
| system | ARM Linux （Debian 12） |
| CPU | Eight-core processor with a maximum clock speed of 2.2GHz |
| GPU | Mali-G52 MC3 |
| NPU | 6 Tops AI Computing Power |
| memory | Standard configuration: 8GB LPDDR4 RAM, 64GB storage |
| Display interface | 1 HDMI 2.1 port, supports up to 8K output |
| interface | 4 USB2.0 ports, 2 USB3.0 ports, and 1 USB OTG port |
| 2 RS232 serial ports (RS485 optional) | |
| network | 1 RJ45 port, 1 WIFI port, optional 4G full-network support |
4. Physical Images
5. AI Algorithm
Supports multiple AI algorithms preloaded and customizes algorithms based on application scenarios.
standard algorithm set | Personnel: Crowd detection, unauthorized access detection, climbing detection, off-duty detection, monitoring room absence detection, elevator entrapment detection, entry/exit detection, sleeping on duty detection, smoking detection, and various behavior analysis |
Vehicles: identification of dump trucks, electric vehicles entering elevators, vehicles driving against traffic, illegal parking of non-motorized vehicles, illegal parking of vehicles, vehicles crossing roads, heavy machinery, and separation of pedestrians from vehicles | |
Object: Detection of exposed soil piles, identification of work clothes and uniforms, recognition of reflective clothing, mask recognition, electric bicycle helmet recognition, insulation glove recognition, safety helmet recognition, fire extinguisher placement detection, epidemic prevention clothing recognition, cat and dog recognition | |
Environment: Roadside waste recognition, flame recognition, flame detection, waste overflow recognition, corridor clutter recognition, open equipment room/distribution room/unit door recognition, street vending recognition, smoke recognition, smoke detection | |
custom algorithm | Industrial inspection, production process monitoring, and quality control; |
Collect at least 50,000 samples, including annotation, design, modeling, training, parameter tuning, cropping, adaptation, and iteration |
6.application function
7.Customized development process for AI vision algorithms based on deep learning:
7.1. Label the images/videos from on-site sampling;
7.2、Get the dataset, then start designing the model, training, and more
7.3、After the model is generated, it is converted into an RKNN model based on the RKNPU chip.
8. Application Scenarios
Smart Community/Industrial Park
Application of the algorithm and product: Occupying the fire lane parking, area intrusion identification, fall identification, smoke and open flame identification, boundary detection, occupation of the road, human figure counting, elevator and electric car detection, etc. are widely used in the identification of human and vehicle entry and exit identity verification, vehicle illegal parking, environmental sanitation detection and other violations and uncivilized behavior, fire hazard detection, etc.
area intrusion recognition
fall detection
boundary crossing detection
Elevator battery car inspection
Algorithm and Product Applications: These include firework detection, hazardous zone identification, wandering recognition, human figure detection, personnel presence monitoring, illegal fishing/stranger intrusion detection, fall recognition, and license plate recognition. The solutions are widely implemented in villas, courtyards, fish ponds, farms, aquaculture bases, and unmanned retail outlets.
hanging walk recognition
human figure detection
Food and beverage outlets
Algorithm and product application: Passenger flow statistics, VIP customer identification, smoking detection, phone call recognition, rodent infestation detection, chef's uniform and hat inspection, open flame off-duty, personnel operation standard, smoke and open flame identification, trash bin overflow detection, etc. are widely used in schools, catering, hospitals, etc.
Passenger flow statistics
smoking recognition
Call Detection
Rodent Detection
Smart Campus:
Student Behavior Analysis:
Fall detection, running detection, climbing detection, fight detection, smoke and fire detection, smoking detection, mobile phone detection, companion line detection, and area detection.
