Use AutoTVM and uTVM to optimize ML workloads on embedded devices & microcontrollers

MicroTVM example

We are seeing a massive increase in resource-constraints for embedded devices due to a lack of mature software stacks. With the increase in open-source hardware, the available software support takes a considerable amount of time to develop AI/ML/DL applications. Some of the challenges faced today are that bare-metal devices do not have on-device memory management, and they do not have LLVM support. They are also hard to debug because of rigid programming and cross-compilation interfaces. Due to this, “optimizing and deploying machine learning workloads to bare-metal devices today is difficult”. To tackle these challenges, there have been developments to support TVM, an open-source machine learning compiler framework for CPUs, GPUs, and machine learning accelerators, on these bare-metal devices, and Apache TVM is running an open-source foundation to make this easy. “µTVM is a component of TVM that brings broad framework support, powerful compiler middleware, and flexible autotuning and compilation capabilities […]

SolidRun launches i.MX 8M Plus SOM and devkit for AI/ML applications

SolidRun already offers NXP based solutions with AI accelerators through products such as SolidRun i.MX 8M Mini SoM with Gyrfalcon Lightspeeur 2803S AI accelerator, or Janux GS31 Edge AI server with NXP LX2160A networking SoC, various i.MX 8M SoCs and up to 128 Gyrfalcon accelerators. All those solutions are based on one or more external Gyrfalcon AI chips, but earlier this year, NXP introduced i.MX 8M Plus SoC with a built-in 2.3 TOPS neural processing unit (NPU), and now SolidRun has just unveiled the SolidRun i.MX 8M Plus SoM with the processor together with development kits based on HummingBoard carrier boards. Specifications: SoC – NXP i.MX 8M Plus Dual or Quad with dual or quad-core Arm Cortex-A53 processor @1.6 GHz (industrial) / 1.8 GHz (commercial), with Arm Cortex-M7 up to 800MHz, Vivante GC7000UL 3G GPU (Vulkan, OpenGL ES 3.1, OpenCL 1.2), 2.3 TOPS NPU, 1080p60 H.264/H.265 video encoder, 1080p60 video […]

Qualcomm QCS610 micro SoM and devkit to power AI and ML smart cameras

Qualcomm QCS610 Development Board

Last July, we missed Qualcomm’s announcement of QCS410 and QCS610 processors designed to bring “premium camera technology, including powerful artificial intelligence and machine learning features formerly only available to high-end devices, into mid-tier camera segments”. The new SoC’s were recently brought to our attention by Lantronix as they have just introduced a new Open-Q 610 micro system-on-module (μSOM) based on Qualcomm QCS610 processor, as well as a development kit designed to bring such smart cameras to market. I first got a bit confused by the product name, but this goes without saying that it is completely unrelated to Qualcomm Snapdragon 610 announced over six years ago. Open-Q 610 micro system-on-module Open-Q 610 specifications: SoC – Qualcomm QCS610 CPU – Octa-core processor with 2x Kryo 460 Gold cores @ 2.2 GHz (Cortex-A76 class), and 6x Kryo 430 Silver low-power cores @ 1.8GHz (Cortex-A55 class) GPU – Qualcomm Adreno 612 GPU @ […]

Google Coral Dev Board mini SBC is now available for $100

Buy Coral Dev Board Mini

Google Coral SBC was the first development board with Google Edge TPU. The AI accelerator was combined with an NXP i.MX 8M quad-core Arm Cortex-A53 processor and 1GB RAM to provide an all-in-all AI edge computing platform. It launched for $175, and now still retails for $160 which may not be affordable to students and hobbyists. Google announced a new model called Coral Dev Board Mini last January, and the good news is that the board is now available for pre-order for just under $100 on Seeed Studio with shipping scheduled to start by the end of the month. Coral Dev Board Mini specifications haven’t changed much since the original announcement, but we know a few more details: SoC – MediaTek MT8167S quad-core Arm Cortex-A35 processor @ 1.3 GHz with Imagination PowerVR GE8300 GPU AI/ML accelerator – Google Edge TPU coprocessor with up to 4 TOPS as part of Coral […]

Microchip SAMD21 Machine Learning Evaluation Kits Work with Cartesiam, Edge Impulse and Motion Gestures Solutions

SAMD21 Machine Learning

While it all started in the cloud Artificial Intelligence is now moving at the very edge is ultra-low power nodes, and Microchip has launched two SAMD21 Arm Cortex-M0+ machine learning evaluation kits that now work with AI/ML solutions from Cartesiam, Edge Impulse, and Motion Gestures. Bot machine learning evaluation kits come with SAMD21G18 Arm Cortex-M0+  32-bit MCU, an on-board debugger (nEDBG), an ATECC608A CryptoAuthentication secure element, ATWINC1510 Wi-Fi network controller, as well as Microchip MCP9808 high accuracy temperature sensor and a light sensor. But EV45Y33A development kit is equipped with an add-on board featuring Bosch’s BMI160 low-power Inertial Measurement Unit (IMU), while EV18H79A features an add-on board with TDK InvenSense ICM-42688-P  6-axis MEMS. The photo above makes it clear both SAMD21 machine learning evaluation kits rely on the same baseboard with a MikroBus socket connected to either 6DOF IMU 2click or 6DOF IMU 14 click add-on board from MikroElektronika. Both […]

CrowPi2 Raspberry Pi 4 Education Laptop Review

CrowPi2 Raspberry Pi Laptop Review

I started my review of CrowPi2 Raspberry Pi 4 Learning Kit a while ago and at the time I showed content from the kit and its first boot. I’ve now spent more time with this very special Raspberry Pi 4 laptop and will focus this review on the education part, namely CrowPi2 software, but will also look at thermal cooling under stress with and without a fan, and try to install another Raspberry Pi compatible board inside the laptop shell. CrowPi2 Education Software It’s quite important to read the user manual before getting started as there are a few non-intuitive steps you may have to take. First I assume the wireless keyboard would just connect after pressing the power button, but it did not. The user manual explains the RF dongle is inside the mouse, and once you connect it you’ll be able to use the keyboard that has some […]

Edge Impulse Enables Machine Learning on Cortex-M Embedded Devices

Edge Impulse

Artificial intelligence used to happen almost exclusively in the cloud, but this introduces delays (latency) for the users and higher costs for the provider, so it’s now very common to have on-device AI on mobile phones or other systems powered by application processors. But recently there’s been a push to bring machine learning capabilities to even lower-end embedded systems powered by microcontrollers, as we’ve seen with GAP8 RISC-V IoT processor or Arm Cortex-M55 core and the Ethos-U55 micro NPU for Cortex-M microcontrollers, as well as Tensorflow Lite. Edge Impulse is another solution that aims to ease deployment of machine learning applications on Cortex-M embedded devices (aka Embedded ML or TinyML) by collecting real-world sensor data, training ML models on this data in the cloud, and then deploying the model back to the embedded device. The company collaborated with Arduino and announced support for the Arduino Nano 33 BLE Sense and […]

ECM3532 AI Sensor Board Features Cortex-M3 MCU & 16-bit DSP “TENSAI” SoC for TinyML Applications

ECM3532 AI Sensor Board

Eta Compute ECM3532 is a system-on-chip (SoC) with a Cortex-M3 microcontroller clocked at up to 100 Mhz, and NXP CoolFlux 16-bit DSP designed for machine learning on embedded devices, aka TinyML, and part of the company’s TENSAI platform. The chip is also integrated into the ECM3532 AI sensor board featuring two MEMS microphones, a pressure & temperature sensor, and a 6-axis motion sensor (accel/gyro) all powered by a CR2032 coin-cell battery. ECM3532 AI sensor board specifications: SoC – ECM3532 neural sensor processor with Arm Cortex-M3 core @ up to 100 MHz (< 5μA/MHz run mode) combines with 512KB embedded FLASH, 256KB SRAM, and 8KB BootROM + secure bootloader, and NXP CoolFlux 16-bit DSP @ up to 100 MHz with 32KB program memory, 64KB data memory. See the product brief for details. Storage – 64Mbit serial Flash for datalogging Connectivity – Bluetooth 4.2 LE via ABOV Semiconductor A31R118 and PCB antenna […]

Exit mobile version