Sipeed MAix: AI at the edge
AI is pervasive today, from consumer to enterprise applications. With the explosive growth of connected devices, combined with a demand for privacy/confidentiality, low latency and bandwidth constraints, AI models trained in the cloud increasingly need to be run at the edge.
MAIX is Sipeed’s purpose-built module designed to run AI at the edge, we called it AIoT. It delivers high performance in a small physical and power footprint, enabling the deployment of high-accuracy AI at the edge, and the competitive price make it possible embed to any IoT devices. As you see, Sipeed MAIX is quite like Google edge TPU, but it act as master controller, not an accelerator like edge TPU, so it is more low cost and low power than AP+edge TPU solution.
MAix's Advantage and Usage Scenarios:
Inherit the advantage of K210's small footprint, Sipeed MAIX-I module, or called M1, integrate K210, 3-channel DC-DC power, 8MB/16MB/128MB Flash (M1w module add wifi chip esp8285 on it) into Square Inch Module. All usable IO breaks out as 1.27mm(50mil) pins, and pin's voltage is selectable from 3.3V and 1.8V.
Firstly, We make an prototype development board for M1, called M1 dock or Dan Dock, it is simple, small, cheap, but all functions include.
As many DIYer want build their own work with breadboard, Sipeed newly provide breadboard-friendly board for you, it called MAix BiT
And the bigger and better MAIX's development board Sipeed MAIX Go
MAix Go is bigger and better than M1 Dock.
MAIX's peripheral module
We have see basic camera and LCD interface on board, so DVP Camera and MCU LCD is support very early.
It support OV7725， OV2640(default, 2M)，OV5640 for the moment.And it support 2.4 inch st7789, and 2.8 inch ili9341 LCD in QVGA resolution.
We have extra tiny I2S mic module and Cool 6+1 Microphone Array Module
Sipeed 6+1 Microphone Arra is a 6 microphone expansion board for Maix AI development boards designed for AI and voice applications.
Including 6+1 digital microphones, 12 three-color LEDs, it supports sound localization, beam forming, speech recognition etc.
Next one, we have Binocular camera module!You can try binocular stereo vision with it
Sipeed Binocular camera module is a camera expansion board for Maix AI development boards designed for AI and Binocular stereo vision applications.
It supports Binocular stereo vision and depth vision. The camera(OV2640 or OV7725) and I2S mic are optional. MAix's SoftWare
MAIX support original standalone SDK, FreeRTOS SDK base on C/C++.
And we port micropython on it: http://en.maixpy.sipeed.com/. It support FPIOA, GPIO, TIMER, PWM, Flash, OV2640, LCD, etc. And it have zmodem, vi, SPIFFS on it, you can edit python directly or sz/rz file to board. We are glad to see you contribute for it:
https://github.com/sipeed/MaixPy //Maixpy project
https://github.com/sipeed/MaixPy_Doc_Us_En_Backup //Maixpy wiki project
MAix's Deep learning
MAIX support fixed-point model that the mainstream training framework trains, according to specific restriction rules, and have model compiler to compile models to its own model format.
It support tiny-yolo, mobilenet-v1, and, TensorFlow Lite! Many TensorFlow Lite model can be compiled and run on MAIX! And We will soon release model shop, you can trade your model on it.
|CPU : RISC-V Dual Core 64bit, 400Mh adjustable||Powerful dual-core 64-bit open architecture-based
processor with rich community resources
|FPU Specifications||IEEE754-2008 compliant high-performance pipelined FPU|
|Debugging Support||High-speed UART and JTAG interface for debugging|
|Neural Network Processor (KPU)||• Supports the fixed-point model that the mainstream training framework trains according to specific restriction rules
• There is no direct limit on the number of network layers, and each layer of convolutional neural network parameters can be configured separately, includ- ing the number of input and output channels, and the input and output line width and column height
• Support for 1x1 and 3x3 convolution kernels
• Support for any form of activation function
• The maximum supported neural network parameter size for real-time work is 5MiB to 5.9MiB
• The maximum supported network parameter size when
working in non-real time is (flash size - software size)
|Audio Processor (APU)||• Up to 8 channels of audio input data, ie 4 stereo channels
• Simultaneous scanning pre-processing and beamforming for sound sources in up to 16 directions
• Supports one active voice stream output
• 16-bit wide internal audio signal processing
• Support for 12-bit, 16-bit, 24-bit, and 32-bit input data widths • Multi-channel direct raw signal output
• Up to 192kHz sample rate
• Built-in FFT unit supports 512-point FFT of audio data
•Uses system DMAC to store output data in system memory
|Static Random-Access Memory (SRAM)||The SRAM is split into two parts, 6MiB of on-chip
general-purpose SRAM memory and 2MiB of on-chip AI SRAM memory, for a total of 8MiB
|Field Programmable IO Array (FPIOA/IOMUX)||FPIOA allows users to map 255 internal functions to 48
free IOs on the chip
|Digital Video Port (DVP)||Maximum frame size 640x480|
|FFT Accelerator||The FFT accelerator is a hardware implementation of the
Fast Fourier Transform (FFT)
|FreeRtos & Standard SDK||Support FreeRtos and Standrad development kit.|
|MicroPython Support||Support MicroPython on M1|
|Machine vision||Machine vision based on convolutional neural network|
|Machine hearing||High performance microphone array processor|
|Supply voltage of external power supply||5.0V ±0.2V|
|Supply current of external power supply||>300mA|
|Range of working temperature||-30℃ ~ 85℃|
|Dimensions||25mm x25mm x1mm|