Edge Face Recognition Project-Union Image/unimage/AIoT Visual Solution Provider

Home

Store

Recently added:

Items 0 Total USD 0

Nickname | quit

The computing power chip used in this project is the RDK X5 Module, the sensor is undetermined, and it has an 8-megapixel wide-angle camera

CPU 8x A55 @1.5GHz

32 Gflops

10 TOPS, supports int8 quantization

4GB/8GB LPDDR4

Supports up to 4K H.264/H.265 60fps video encoding

Supports 2-channel 4-wire MIPI-CSl interface

Supports HDMI interface, up to 1080p@60fps

Supports 1-channel 4-line MIPI-DSl interface

CPU:Single-core Cortex-A7 + MCU

Video Output Interface:Support SPI interface display

CPU:2D Graphics Engine

ISP:8M@20fps

NPU:0.5 TOPS, supporting int8 quantization

Audio Interface and Processing:Integrated Audio codec, 1*24-bit ADC and 24-bit DAC

Storage Interface:Built-in 512Mb ~ 2Gb DDR, eMMC 4.51

Peripheral Interface:USB2.0/ 3xUART/ SPI/ 5xI2C/ 12xPWM/ 2xSDIO3.0/ SARADC

Video Codec:Highest support for 4K H.264/H.265 25fps video encoding

Packaging:OFN88 9*9mm

Video Input Interface:Supports 2-channel 4-lane or 4-channel 2-lane MIPI-CSI interfaces

Software Features

Stable Multi-Object Tracking

Identification under backlight, night-time supplementary lighting, and Supports multi-person tracking in dynamic scenarios with 10–20 people, short-term occlusion recovery time ≤1s, and trajectory retention rate ≥80% in multi-person intersection scenarios

Algorithm Voting and Trajectory-Level Fusion

Identification under backlight, night-time supplementary lighting, and Adopt the architecture of Detection → Tracking → Quality Assessment → ROI Enhancement → Triggered Recognition → Trajectory-Level Fusion, and prevent false recognition through quality gating and algorithm voting mechanisms

Dynamic Real-time Recognition

Identification under backlight, night-time supplementary lighting, and Supports recognition during movement (stationary, slow walking, running, skipping rope, turning head, partial occlusion), with the end-side dynamic processing frame rate ≥15 FPS and the static first confirmation latency <3s

Long-distance small face enhancement

Identification under backlight, night-time supplementary lighting, and Addressing the issues of recognition distance of 8–10m and edge distortion of wide-angle lenses, it supports ROI enhancement and zoning strategies, with a face recall rate of ≥85% in the edge area

Efficient Deployment of Edge-side Quantization

Identification under backlight, night-time supplementary lighting, and Based on Horizon X5 BPU's 10 TOPS computing power, the model is deployed with edge-side quantization, peak resource utilization is controlled, and it operates stably for 7×12h continuously

Standardized Platform Interface

Identification under backlight, night-time supplementary lighting, and Supports standardized interface docking with upper-level student face registration library/device management platform/business application platform, and supports full/incremental synchronization and encrypted template storage

Hardware Features

High-performance Edge AI Chip

BPU Computing Power 10 TOPS, 8-core Cortex-A55 @ 1.5GHz, Supports Efficient Inference of Edge-side Quantized Models

Large-capacity storage configuration

Equipped with 2GB/4GB LPDDR4 memory + 8GB/16GB eMMC storage to meet the needs of multi-template caching and local event supplementary transmission

Wide Field of View Coverage

Horizontal Field of View H: 112.1°(W)–47.5°(T), Vertical Field of View V: 58.0°(W)–26.6°(T), supports wide-angle fixed-focus/zoom lenses, covering a large monitoring area

Multi-channel Video Interface

2 × 4-lane MIPI CSI video input, 1 × HDMI (up to 1080p@60fps) + 1 × 4-lane MIPI DSI video output

Enrich Peripheral Interface

1 × USB 3.0 Host、1 × USB 2.0 Device、1 × Debug UART、6 × UART、4 × SPI、7 × IIC、1 × SDIO、8 × LPWM、8 × PWM

Dual-mode network communication

1 × Gigabit Ethernet + Wi-Fi 6 / Bluetooth 5.4, supporting flexible wired/wireless deployment

To become a global leader in AloT vision system solution

Co-creating a smarter future across industries through inneovative AloT vision systems

About Us Contact Us

Location:Building A2, Dingbao Hong Green Hi-Tech Park, Shiyan Street, Bao'an District, Shenzhen, China.

Email:global@unionimage.com

Tel: