Raspberry Pi Pico W for Tiny Machine Learning:
Raspberry Pi Pico is a microcontroller which has dual core Arm Cortex M0+ architecture manufactured and distributed by Raspberry Pi. It has 264KB of RAM but no flash memory due to space and cost constraints. The Pico W board is featured with an external flash memory of 4MB. But this RAM and Flash memory is sufficient for many small edge machine learning applications like Keyword spotting, Vibration analysis through accelerometer , Gesture recognitions and similar applications which can be done in Edge Impulse online tool.
Edge impulse tool is capable of acquiring data from the sensors through microcontrollers (eg. Pico W) and collect them in their database. Then we need to choose step by step process to generate Neural network model from the dataset and deploy the model file to the microcontroller as a library in C or binary file . If the model file is generated as a Arduino library , the user code can be utilized according to the user's needs or application. Else the user can directly use the application running from the microcontroller or derive and observe the Neural network output in serial console window (in Arduino serial window or software like Putty).
The user need to connect a sensor like PDM Microphone module or a Accelerometer sensor for obtaining the data and training the neural network. Pi Pico can be easily connected to the modules like microphone and accelerometer making the system very cheaper. Other analog sensor and Ultrasonic Distance sensor ca also be used for some NN applications . Pi Pico also comes as a Readymade board named Arduino Nano RP2040 connect from Arduino which has onboard Accelerometer and PDM microphone. For more details on the TinyML documentation in Edge Impulse, check the link given here.
Seeed Studio Xiao Boards:
Seeedstudio is manufacturing a range of XIAO series microcontroller boards based on ESP32, RP2040 Pico , SAMD51 and NRF52840. These boards are less cost and can be used for various applications like IOT and Machine Learning. In particular two boards in XIAO series are used for machine learning again supported by Edge Impulse are XIAO ESP32S3 sense and NRF52840 sense. Both these boards have on-board 3.7V Lithium battery charging capability from USB.
Seeed Studio XIAO ESP32S3 Sense:
This board has a ESP32S3 processor and has 2.4GHz Wi-Fi, BLE 5.0, OV2640 camera sensor, onboard digital microphone, 8MB PSRAM, 8MB FLASH, battery charge supported, rich Interface, IoT, embedded ML. This board can perform machine learning since it has high RAM in platforms like Edge impulse and Seeedstudio's SensecraftAI. This board can be used for applications like Keyword spotting , Gesture recognition.
Seeed Studio XIAO nRF52840 Sense:
This board equipped with a powerful Nordic nRF52840 MCU which integrates Bluetooth 5.0 connectivity. This board also has 6 axis PDM microphone and 6 axis IMU (Accelerometer and Gyroscope). The board supports Nearfield communication (NFC) which is used for easier Bluetooth connection establishment with NFC featuring mobile phones. This board can be used for applications like Keyword spotting , Gesture recognition using SensecraftAI AI and Edge Impulse online tools.
Grove Vision AI Module V2
It is an MCU-based vision AI module powered by WiseEye2 HX6538 Arm Cortex-M55 & Ethos-U55 processor . It supports TensorFlow and PyTorch frameworks and is compatible with Arduino IDE. AI models can be trained in the device using SensecraftAI and Edge impulse tools. It features a standard camera CSI interface, an onboard digital microphone and an SD card slot, making it highly suitable for various embedded AI vision projects. This module can be interfaced with various Arduino , Raspberry Pi and Xiao boards for Machine learning and Vision based applications easily. For machine vision based applications , OV5647-62 FOV Camera Module for Raspberry Pi is connected with this module.