How’s the performance of Grove Vision AI V2?
To have a quick evaluation of this board’s performance, we compared it to Seeed’s other MCU-based vision AI boards – Grove Vision AI Module, XIAO ESP32S3 Sense in four areas:
1. Power consumption: This metric indicates whether the board can be used in battery-powered products.
2. Inference time: This metric indicates the processing speed of the MCU and how much latency is involved.
3. Frame rate: This metric evaluates whether the product can capture instant changes, patterns, and movements.
4. Ease of use: We assessed whether this product is user-friendly for vision AI novices and can quickly run mainstream models on the market.
Based on the evaluation results, the Grove Vision AI V2 boasts an impressive inference time of just 33 milliseconds, a fast frame rate of 30.30 FPS, a gaming like smooth experience, and efficient power consumption at only 0.35 watts. These results clearly demonstrate that the Grove Vision AI V2 is a premium choice for high-performance vision processing.
What’s on the board?
Grove Vision AI V2 is now compatible with all Pi cameras through a standard CSI interface.
Grove Vision AI V2 is not only designed for vision applications but also features an onboard PDM microphone for sound applications. It comes with a SD card slot allows for convenient storage of images, videos, and identification results using an SD card.
With various interfaces like IIC, UART, SPI, and Type-C, this board has expansive capabilities and can be easily connected to popular products such as Seeed Studio XIAO, Grove, Raspberry Pi, BeagleBoard and ESP-based products for further development. For instance, integrating Grove Vision AI V2 with XIAO can effortlessly access the interface and data of Grove Vision AI V2 through Arduino, Micropython, CircuitPython, and PlatformIO, and conveniently connect to the cloud or dedicated servers like Home Assistance.
Applications
Earlier MCU-based vision sensors were constrained by limited computing power, typically only supporting simple tasks like face detection. We believe this product surpasses the limits, enabling functionalities that were previously unachievable with traditional MCU-based vision AI sensors, such as:
1. Facial Keypoint Detection: Beyond basic facial recognition, detailed facial keypoint detection is capable of running. It accurately identifies critical features such as eyes, nose, and mouth.
2. Human Pose Estimation: It analyzes key joints and limbs to determine human posture, identifying body parts like arms, legs, torso, and head. This is useful for fitness tracking, posture correction, and motion recognition.
3.Object Detection: It supports powerful models like the Yolo v8n model, which can detect and classify over 80 types of objects.
Part List
| Grove Vision AI V2 |
x1 |
| Cable |
x1 |
ECCN/HTS
| HSCODE |
9031900090 |
| USHSCODE |
8517180050 |
| UPC |
|
| EUHSCODE |
9013101000 |
| COO |
CHINA |
|
Microcontroller
|
Himax WiseEye2 HX6538 processor featuring a dual-core Arm Cortex-M55 and integrated Arm Ethos-U55
|
|
Onboard Peripherals
|
PDM Microphone, SD Card Slot
|
|
Rich Interfaces
|
CSI, IIC, UART, SPI, and Type-C
|
|
Input Voltage
|
5V
|
|
Power Supply
|
Dual 7-pin connector & Type-C & Grove Connector
|
|
Rate
|
115200
|
|
I2C Interface
|
Seeed Studio XIAO & Arduino
|
|
Downloading & Firmware Burn Interface
|
Type-C
|
|
Frequency(ARM Cortex-M55 Processor(Big))
|
Up to 400MHz
|
|
Frequency(ARM Cortex-M55 Processor(Little))
|
Up to 150MHz
|
|
Frequency(ARM Ethos-U55 MicroNPU)
|
Up to 400MHz
|
|
Memory Card Interface
|
Up to 1x SD and SDIO host, support DS mode, up to 25MHz
|
|
Internal System Memory
|
● Configurable system memory, up to 2432KB
● 64KB boot ROM
|