machine learning News - CNX Software

March 13, 2025 by Debashis Das - No Comments

Renesas RZ/V2N low-power AI MPU integrates up to 15 TOPS AI power, Mali-C55 ISP, dual MIPI camera support

Renesas has recently introduced the RZ/V2N low-power Arm Cortex-A55/M33 microprocessor designed for machine learning (ML) and computer vision applications. It features the company’s DRP-AI3 coprocessor, delivering up to 15 TOPS of INT8 “pruned” compute performance at 10 TOPS/W efficiency, making it a lower-power alternative to the RZ/V2H. Built for mid-range AI workloads, it includes four Arm Cortex-A55 cores (1.8GHz), a Cortex-M33 sub-CPU (200MHz), an optional 4K image signal processor, H.264/H.265 hardware codecs, an optional Mali-G31 GPU, and a dual-channel four-lane MIPI CSI-2 interface. The chip is around 38% smaller than the RZ/V2H MPU and operates without active cooling. The RZ/V2N is suitable for applications like endpoint vision AI, robotics, and industrial automation. Renesas RZ/V2N specifications CPU Application Processor – Quad-core Arm Cortex-A55 @ 1.8 GHz (0.9V) / 1.1 GHz (0.8V) L1 cache – 32KB I-cache (with parity) + 32KB D-cache (with ECC) per core L3 cache – 1MB (with ECC, […]

March 6, 2025 by Jean-Luc Aufranc (CNXSoft) - No Comments

Zant – An open-source Zig SDK for neural network deployment on microcontrollers

Zant neural network deployment microcontrollers

Zant is an open-source, cross-platform SDK written in Zig and designed to simplify deploying Neural Networks (NN) on microcontrollers. It comprises a suite of tools to import, optimize, and deploy NNs to low-end hardware. The developers behind the project developed Zant (formerly known as Zig-ant) after noticing many microcontrollers lacked robust deep learning libraries, and made sure it would be on various platforms such as ARM Cortex-M or RISC-V microcontrollers, or even x86 targets. Contrary to platforms like Edge Impulse that focus on network creation, Zant is about deployment and outputs a static, highly optimized library ready to be integrated into any existing work stack. Zant highlights: Optimized Performance – Supports quantization, pruning, and hardware acceleration techniques such as SIMD and GPU offloading. Low memory footprint – Zant employs memory pooling, static allocation, and buffer optimization to work on resources-constrained targets. Ease of Integration: With a modular design, clear APIs, […]

February 27, 2025February 27, 2025 by Jean-Luc Aufranc (CNXSoft) - 2 Comments

Arm Cortex-A320 low-power CPU is the smallest Armv9 core, optimized for Edge AI and IoT SoCs

Arm Cortex-A320 is a low-power Armv9 CPU core optimized for Edge AI and IoT applications, with up to 50% efficiency improvements over the Cortex-A520 CPU core. It is the smallest Armv9 core unveiled so far. The Armv9 architecture was first introduced in 2021 with a focus on AI and specialized cores, followed by the first Armv9 cores – Cortex-A510, Cortex-A710, Cortex-X2 – unveiled later that year and targeting flagship mobile devices. Since then we’ve seen Armv9 cores on a wider range of smartphones, high-end Armv9 motherboards, and TV boxes, The upcoming Rockchip RK3688 AIoT SoC also features Armv9 but targets high-end applications. The new Arm Cortex-A320 will expand Armv9 usage to a much wider range of IoT devices including power-constrained Edge AI devices. Arm Cortex-A320 highlights: Architecture – Armv9.2-A (Harvard) Extensions Up to Armv8.7 extensions QARMA3 extensions SVE2 extensions Memory Tagging Extensions (MTE) (including Asymmetric MTE) Cryptography extensions RAS extensions […]

February 20, 2025February 20, 2025 by Debashis Das - 2 Comments

LLMStick – An AI and LLM USB device based on Raspberry Pi Zero W and optimized llama.cpp

Youtuber and tech enthusiast Binh Pham has recently built a portable plug-and-play AI and LLM device housed in a USB stick called the LLMStick and built around a Raspberry Pi Zero W. This device portrays the concept of a local plug-and-play LLM which you can use without the internet. After DeepSeek shook the world with its performance and open-source accessibility, we have seen tools like Exo that allow you to run large language models (LLMs) on a cluster of devices, like computers, smartphones, and single-board computers, effectively distributing the processing load. We have also seen Radxa release instructions to run DeepSeek R1 (Qwen2 1.5B) on a Rockchip RK3588-based SBC with 6 TOPS NPU. Pham thought of using the llama.cpp project as it’s specifically designed for devices with limited resources. However, running llama.cpp on the Raspberry Pi Zero W wasn’t straightforward and he had to face architecture incompatibility as the old […]

February 17, 2025February 18, 2025 by Debashis Das - No Comments

Waveshare ESP32 robotic arm kit with 5+1 DoF supports ROS2, LeRobot, and Jetson Orin NX integration

RoArm M3 Pro and RoArm M3 S High Torque Serial Bus Servo Robotic Arm Kit

Waveshare has recently released the RoArm-M3-Pro and RoArm-M3-S, a 5+1 DOF high-torque ESP32 robotic arm kit. The main difference between the two is that the RoArm-M3-Pro features all-metal ST3235 bus servos for durability and longevity, on the other hand, the RoArm-M3-S uses standard servo motors which are less durable for long-term use. These robotic arms feature a lightweight structure, a 360° omnidirectional base, and five flexible joints, which together create a 1m workspace with a 200 grams @ 0.5m payload. A 2-DOF wrist joint enables multi-dimensional clamping and precise force control. It integrates an ESP32 MCU, supporting multiple wireless control modes via a web app, it also supports inverse kinematics for accurate positioning, curve velocity control for smooth motion, and adaptive force control. The design is open source and with ROS2 compatibility, it allows secondary development via JSON commands and ESP-NOW for multi-device communication. Compatible with the LeRobot AI framework, […]

February 7, 2025February 7, 2025 by Jean-Luc Aufranc (CNXSoft) - No Comments

FOSSASIA 2025 – Operating systems, open hardware, and firmware sessions

The FOSSASIA Summit is the closest we have to FOSDEM in Asia. It’s a free and open-source event taking place each year in Asia, and FOSSASIA 2025 will take place in Bangkok, Thailand on March 13-15 this year. It won’t have quite as many speakers and sessions as in FOSDEM 2025 (968 speakers, 930 events), but the 3-day event will still have over 170 speakers and more than 200 sessions. Most of the sessions are for high-level software with topics like AI and data science, databases, cloud, and web3, but I also noticed a few sessions related to “Hardware and firmware” and “Operating System” which are closer to what we cover here at CNX Software. So I’ll make a virtual schedule based on those two tracks to check out any potentially interesting talks. None of those sessions take place on March 13, so we’ll only have a schedule for March […]

February 6, 2025February 7, 2025 by Jean-Luc Aufranc (CNXSoft) - 2 Comments

YOLO-Jevois leverages YOLO-World to enable open-vocabulary object detection at runtime, no dataset or training needed

YOLO-Jevois general object detection by typing words

YOLO is one of the most popular edge AI computer vision models that detects multiple objects and works out of the box for the objects for which it has been trained on. But adding another object would typically involve a lot of work as you’d need to collect a dataset, manually annotate the objects you want to detect, train the network, and then possibly quantize it for edge deployment on an AI accelerator. This is basically true for all computer vision models, and we’ve already seen Edge Impulse facilitate the annotation process using GPT-4o and NVIDIA TAO to train TinyML models for microcontrollers. However, researchers at jevois.org have managed to do something even more impressive with YOLO-Jevois “open-vocabulary object detection”, based on Tencent AI Lab’s YOLO-World, to add new objects in YOLO at runtime by simply typing words or selecting part of the image. It also updates class definitions on […]

January 31, 2025January 31, 2025 by Jean-Luc Aufranc (CNXSoft) - 1 Comment

Silicon Labs unveils low-cost BG22L BLE 5.4 and BG24L BLE 6.0 SoCs

Silicon Labs has unveiled the BG22L and BG24L SoCs for low-cost, ultra-low-power Bluetooth LE connectivity. These are Lite versions of the company’s BG22 and BG24 SoC families introduced in 2020 and 2022 respectively. The 38.4 MHz Silabs BG22L Arm Cortex-M33 SoC targets high-volume, cost-sensitive Bluetooth 5.4 applications like asset tracking tags and small appliances. In comparison, the 78 MHz BG24L Cortex-M33 SoC offers an affordable entry-level solution with AI/ML acceleration and Bluetooth 6.0 Channel Sounding to locate items or implement access control in crowded areas such as warehouses and multi-family housing. Since the specs for the BG22L and BG24L are similar to the ones for the BG22 and BG24 chips I won’t reproduce those here, and instead highlights the main features and cost-saving measures. Silicon Labs BG22L Bluetooth LE 5.4 SoC Silicon Labs BG22L highlights: MCU – Arm Cortex-M33 @ 38.4 MHz with DSP and FPU (BG22 is clocked at […]