Embarking on Computer Vision for Facial and Voice Recognition with Low-Cost Artificial Intelligence of Things (AIoT) Modules RISC-V : K210, Sipeed M1 or M1W
Introduction
Computer vision is a rapidly expanding technology that offers endless possibilities. With low-cost Artificial Intelligence of Things (AIoT) modules like the K210, Sipeed M1 RISC-V or M1W, you can now create facial and voice recognition applications without breaking the bank. With deep learning frameworks such as TensorFlow, Keras, and Darknet, these AIoT modules can be programmed with development environments such as MaixPy IDE, PlatformIO IDE, OpenMV IDE, Arduino IDE, etc., using programming languages such as C, C++, and MicroPython, you can start working on real projects with Chinese modules purchased from Aliexpress.
There is a wide variety of development tools for K210 boards, Sipeed M1 RISC-V, and Sipeed M1W RISC-V available on AliExpress. Here are some of the low cost module you can find :
Maix Dock (M1W) :
This is an AIoT development board based on Kendryte’s K210 (64-bit dual-core RISC-V architecture). It is equipped with a Type-C port and a USB-UART circuit, allowing users to connect directly to a computer for development.
Sipeed M1n Module AI Development Kit :
This AIoT module development kit is designed by Sipeed. It includes an M1n module, a Type-C to M.2 adapter (not a standard M.2 interface), and a camera.
Sipeed Maixduino AI Development Board :
This AIoT development board is based on the K210 RISC-V.
Maix Bit AI Development Board :
This AIoT development board adopts the K210 AI chip which is equipped with a Neural Processing Unit (NPU) that can perform convolutional neural network operations with high performance.
AI-Motion K210 Developer Kit from Yahboom
The Al-Motion K210 Developer Kit, a compact yet comprehensive development board kit, is a collaborative creation by Yahboom and Canaan. It employs the RISC-V processor architecture in the K210, which is specifically engineered for multi-modal shape recognition in machine vision and machine hearing. This makes it highly applicable in a range of sectors, including smart homes, machine vision, intelligent robotics, and security surveillance.
Features :
- Developed by Yahboom and Canaan.
- Uses the RISC-V processor architecture in the K210, designed for multi-modal shape recognition in machine vision and machine hearing.
- Contains a variety of peripherals such as a capacitive touchscreen, camera, microphone, speaker, six-axis attitude sensor, and a WiFi module.
The development board is equipped with an array of peripherals such as a capacitive touchscreen, camera, microphone, speaker, six-axis attitude sensor, and a WiFi module. In addition to these features, it also offers a wealth of development resources to aid in further development and learning. This significantly reduces the learning curve for AI vision technology.
For more information, you can see the Youtube video or visit the official Yahboom website or purchase it on Aliexpress
Sipeed M1 RISC-V Module
The Sipeed M1 RISC-V module is a powerful and affordable solution for computer vision applications. It is based on the Kendryte K210 dual-core 64-bit RISC-V processor and is capable of handling image and voice recognition tasks.
Features :
- Developed based on Canaan Kanzhi Technology’s edge intelligent computing chip K210 (RISC-V architecture).
- The main control chip has a built-in 64-bit dual-core high-performance low-power processor, each core has a floating-point unit (FPU), a convolutional artificial neural network intelligent hardware accelerator (KPU) and a fast Fourier transform accelerator (FFT).
- Supports a variety of mainstream AI programming frameworks.
- Does not have a WiFi module.
Sipeed M1W RISC-V :
- Similar to the Sipeed M1 RISC-V in terms of AI functionality.
- The main difference is that the M1W module has an additional WiFi functionality, provided by the ESP8285 WiFi chip.
Grove AI HAT for Edge Computing :
The Grove AI HAT for Edge Computing, a cost-effective yet robust Raspberry Pi AI hat, is designed around the Sipeed MAix M1 AI MODULE, which incorporates the Kendryte K210 processor. This device not only enhances the Raspberry Pi’s ability to execute AI at the edge but can also operate independently for edge computing applications.
Use Cases
The AIoT module can be used in various low-cost computer vision applications. For example, it can be used for face detection, object recognition in images and videos, and even for smart agriculture applications.
Here are some specific examples:
- Face Detection : The AioT module can be used to develop a system that detects faces in real-time. This could be used in a variety of settings, such as security systems where access is granted based on facial recognition.
- Object Recognition : The AIoT module can also be used for object recognition in images and videos. For example, it could be used in a retail setting to automatically identify items on the shelves, helping to manage inventory and restocking.
- Smart Agriculture : In the field of agriculture, the AIoT module can be used to develop systems that monitor crop health and detect pests1. This could help farmers to take timely action, improving crop yields and reducing losses.
- Voice Recognition : Beyond image processing, the AIoT module can also handle voice recognition tasks. This could be used to develop low-cost smart home systems where various appliances are controlled through voice commands.
- Gesture Recognition : The module can be used to develop systems that recognize specific gestures. This could be used in interactive installations or in developing more intuitive interfaces for controlling devices.
- Autonomous Vehicles : In the field of robotics, the AIoT module can be used to develop autonomous vehicles. The module can process visual data in real-time, allowing the vehicle to navigate its environment.
- Reading QR Code, Data Matrix, and Bar Code : The AIoT module can be used in retail and logistics industries for scanning and reading QR codes, data matrices, and bar codes. This can help in tracking inventory, validating tickets, or even in mobile payment systems.
- Counting Spare Parts : In manufacturing and production industries, the AIoT module can be used for counting spare parts. It can identify and count specific parts based on their shape, size, and color. This can help in maintaining accurate inventory levels and in identifying when reordering of parts is necessary.
- Quality Control : The AIoT module can also be used for quality control in various industries. By using image recognition and machine learning algorithms, it can identify defects or irregularities in products on a production line. This can help in maintaining high quality standards and reducing waste.
These are just a few examples of the many possible applications of the AIoT module in the field of computer vision. With its affordability and versatility, the AIoT module opens up a world of possibilities for developers and hobbyists alike.
Tools to Use
To develop computer vision projects with AIoT module, several tools can be used :
- OpenCV : An open-source software library for image and video processing.
- TensorFlow : An open-source machine learning tool that can be used for various tasks, including computer vision.
- PyTorch : Another open-source machine learning library that is particularly useful for computer vision applications.
- Keras : A high-level framework for deep learning.
- Caffe : A deep learning framework dedicated to speed and efficiency.
- MaixPy : An easy-to-use development environment for creating AI projects with Python on the AIoT module.
Creating a Complete Project
To create a complete project with the AioT module, you can follow these steps :
- Setting up the development environment : Install the necessary tools for development, such as MaixPy or other tools mentioned above.
- Defining the problem : Identify the problem you want to solve with computer vision.
- Collecting and preparing data : Collect the images or videos needed for your project. You may need to annotate or prepare them in some way for model training.
- Training the model : Use a tool like TensorFlow or PyTorch to train a computer vision model on your data.
- Deploying the model : Once the model is trained, you can deploy it on the AIoT module. You can use MaixPy for this.
- Testing and optimization : Test your project and optimize it if necessary. You may need to adjust your model or collect more data to improve performance.
- Final deployment : Once you are satisfied with the performance of your project, you can deploy it for real use.
Conclusion :
In conclusion, this journey into the domain of computer vision and voice recognition using low-cost artificial intelligence has been both challenging and rewarding. The potential applications of these technologies are vast and can revolutionize many sectors. As we continue to explore and innovate, we are excited to announce that we will soon be sharing projects featuring the AI-Motion K210 Developer Kit from Yahboom. This powerful tool promises to open new horizons in our AI journey. Stay tuned for more updates !
I hope this article will help you in your first computer vision project.
Datasheets and links :
- [Sipeed M1 Datasheet EN V1.12](https://github.com/larbi67/Face-Detection-Sipeed-M1/blob/main/Sipeed%20M1%20Datasheet%20EN%20V1.12.pdf)
- [Sipeed M1W Datasheet EN V1.11](https://github.com/larbi67/Face-Detection-Sipeed-M1/blob/main/Sipeed-M1W-Datasheet-EN-V1.11.pdf)
- [Wiki SIPEED](https://wiki.sipeed.com/en/index.html)
- [SiPEED Official Website](www.sipeed.com)
- [SiPEED Github](https://github.com/sipeed)
- [Sipeed Model Store](https://maixhub.com/welcome)
- [SDK Reference](dl.sipeed.com/MAIX/SDK)
- [HDK Reference](dl.sipeed.com/MAIX/HDK)
- [E-mail Technical Support](support@sipeed.com)
