Grove Vision AI 模組具有一個嵌入邊緣機器學習算法的迷你型人工智能攝像頭，只需單擊即可輕鬆部署。它是各種物體檢測項目的小助手，可以根據不同的需求進行定制。試想一下可通過一個微型感測器賦予設備視覺。
人工智能相機（AI 相機）是由內置邊緣機器學習算法提供支持的增強相機，通過計算攝影進行智能處理，以實時執行增強的對象檢測。它已廣泛用於智能手機的人臉辨識、野生動物檢測的邊緣設備和其他邊緣智能應用。
Grove-Vision AI Module 搭配Himax HX6537-A / OV2640 攝像頭這款迷你型相機感測器將是您開始使用人工智能相機的完美模組。它安裝了一個預訓練模型可讓您檢測人臉。此外它還安裝了三種用於模型辨識的 ML 分類算法，可讓您自由開發用於檢測不同對象的模型並將其直接應用到您的項目中。
該感測器由物聯網硬件支持者 Seeed Studio 推出，包括兩個典型的物聯網模組：一個數字麥克風和一個 6 軸慣性測量單元 (IMU)，供您進一步使用邊緣 AI。
使用 YOLOv5 構建您自己的檢測模型
YOLO 是“You Only Look Once”一詞的縮寫。它是一種實時檢測和識別圖像中各種對象的算法。 Ultralytics YOLOv5 是基於 PyTorch 框架的 YOLO 版本。基於其存儲庫，Seeed Studio 發布了更適合 Seeed Studio AIoT 硬件/設備的版本。您可以使用此存儲庫來構建您自己的檢測模型。它們不僅可以在 Grove 視覺 AI 模組上實現，還可以在工業級 SenseCAP A1101 – LoRaWAN Vision AI Sensor 上實現。
將您的數據傳輸到 SenseCAP 雲
與 SenseCAP 工業級感測器一樣，基礎級 Grove Vision AI Module 可以連接到 SenseCAP 雲進行數據傳輸。為了實現這一功能，我們很自豪地推出 SenseCAP K1100 – 感測器原型套件，具有 LoRa® 和 AI，結合了 Wio Terminal 和其他經典物聯網感測器。
Grove-Vision AI Module 搭配Himax HX6537-A / OV2640 攝像頭套件，使您能夠使用 LoRa® 和 AI 技術快速數字化世界，以應對現實世界的挑戰。有了這個即插即
用的工具包，任何人都可以將 AI 添加到他們的邊緣設備並釋放 AIoT 的潛力。完成簡單易懂的教程後，您可以在 10 分鐘內通過 LoRaWAN® 或 Wi-Fi 將感測器與 SenseCAP 雲連接。
使用更小的MCU開發板建造專案：Seeed Studio XIAO
Grove Vision AI 的介面周圍設有雙 7 針接頭，用於連接 Seeed Studio XIAO 系列開發板。 Seeed Studio XIAO 系列開發板皆為拇指大小，採用SAMD21、nRF52840、RP2040及ESP32C3等熱門高效能晶片。此外，由於其尺寸緊湊，所有 SMD 元件均位於電路板的同一側，因此設計人員可以輕鬆地將 XIAO 整合到電路板中，從而快速生產。
特徵
- 可定制模型：支持導入最多3個定制模型，實現物體檢測等主流ML功能
- 緊湊型 AI 攝像頭：配備 OV2640 SOC 感測器，支持自動曝光和自動白平衡，提供出色的圖像
- 開發部署簡單：支持 Arduino IDE編程 和拖放模型實現
- 豐富的文檔資源：提供詳細的分步指南供用戶入門
產品應用
- 人員檢測
- 自定義對象檢測
- 智慧門衛
- 跌倒檢測和警報
- 人臉辨識
產品規格表
出貨清單
- Grove – Vision AI Module x1
- Grove Cable x1
Documents
Grove-Vision AI Module
Meet the new member of Grove Family: Grove Vision AI Module. The module features a compact AI-powered camera embedded with Edge Machine Learning algorithms and can be easily deployed with simple clicks. It is a small and great assistant for various object detection projects and can be customized based on different needs. Just imagine endowing devices with vision through a tiny sensor.
Key Features
- Customizable Model: Support the import of up to 3 customized models to realize mainstream ML functions like object detection
- Compact AI camera: Equipped with OV2640 SOC sensor that supports auto-exposure and auto-white balance to offer good images
- Simple development & deployment: Support Arduino IDE programming and drag-and-drop model implementation
- Plentiful documentation resource: Provide detailed and step-by-step guides for users to get started
- High compatibility: Perfectly match with XIAO series interface, support Raspberry Pi and Arduino ecology through Grove connector
Artificial Intelligence camera (AI camera) is the enhanced camera powered by a build-in Edge Machine Learning algorithm, smartly processing with computational photography to perform enhanced object detection in real-time. It has been widely used in smartphones for face recognition, edge devices for wildlife detection, and other Edge Intelligence applications.
This compact camera sensor will be the perfect module for you to get started with the AI-powered camera. It is installed with one pre-trained model that allows you to detect human faces. Furthermore, it has been installed with three ML classification algorithms for model recognition, which allows you to freely develop the models for detecting different objects and directly apply them to your projects.
Launched by Seeed Studio, the IoT hardware enabler, the sensor includes two typical IoT modules: a digital microphone and a 6-axis Inertial Measurement Unit (IMU), for your further usage of edge AI. The official software is still under development and will be published in a short term.
Use YOLOv5 to Build Detect Models of Your Own
YOLO is an abbreviation for the term ‘You Only Look Once’. It is an algorithm that detects and recognizes various objects in an image in real-time. Ultralytics YOLOv5 is the version of YOLO based on the PyTorch framework. Based on its repository, Seeed Studio has published a more suitable version that fits in Seeed Studio AIoT hardware/devices. You can use this repository to build detect models of your own. They can not only be implemented on the Grove vision AI module but also on the industrial-level SenseCAP A1101 – LoRaWAN Vision AI Sensor.
Meanwhile, we have provided a specific tutorial to show you how to train the AI model for specific applications and then deploy it onto the devices.
Attention
Please be cautious of the connection method and be careful when you soldered the slot.
Application
- People Detection
- Customized Object Detection
- Smart Doorkeeper
- Fall Detection & Alarm
- Face Recognition
Specification
| Characteristic |
Value |
| Operating Voltage |
5 |
| Rate |
115200 |
| I2C Interface |
Seeed Studio XIAO & Arduino |
| Power Supply |
dual 7-pin connector & Type-C |
| Downloading & Firmware Burn Interface |
Type-C |
| Dimensions |
40 * 20 * 13 |
Hardware Layout
|
Connector |
Grove (Grove base for Arduino) |
5V Charge and Data Transmission |
| Double row 7pin socket (Seeed Studio XIAO) |
5V Charge and Data Transmission |
| USB Type-C |
5V Charge and Firmware Burn |
| Communication Mode |
IIC |
|
| Processor |
Himax HX6537-A |
400Mhz DSP (ultra low power consumption) |
| Camera Sensor |
OV2640 |
Resolution Ratio 1600*1200 |
資料來源：Seeed Studio 官方技術文件，經台灣物聯科技實測整理後彙整
