computer vision

Posted on by Jean-Luc Aufranc (CNXSoft) - 12 Comments on Testing AI and LLM on Rockchip RK3588 using Mixtile Blade 3 SBC with 32GB RAM

Testing AI and LLM on Rockchip RK3588 using Mixtile Blade 3 SBC with 32GB RAM

mixtile blade 3 review RK3588 AI LLM

We were interested in testing artificial intelligence (AI) and specifically large language models (LLM) on Rockchip RK3588 to see how the GPU and NPU could be leveraged to accelerate those and what kind of performance to expect. We had read that LLMs may be computing and memory-intensive, so we looked for a Rockchip RK3588 SBC with 32GB of RAM, and Mixtile – a company that develops hardware solutions for various applications including IoT, AI, and industrial gateways – kindly offered us a sample of their Mixtile Blade 3 pico-ITX SBC with 32 GB of RAM for this purpose. While the review focuses on using the RKNPU2 SDK with computer vision samples running on the 6 TOPS NPU, and a GPU-accelerated LLM test (since the NPU implementation is not ready yet), we also went through an unboxing to check out the hardware and a quick guide showing how to get started […]

Posted on by Jean-Luc Aufranc (CNXSoft) - No Comments on Qualcomm unveils Snapdragon X80 5G modem with NB-NTN satellite connectivity, AI Hub, and FastConnect 7900 WiFi 7 chip

Qualcomm unveils Snapdragon X80 5G modem with NB-NTN satellite connectivity, AI Hub, and FastConnect 7900 WiFi 7 chip

Snapdragon X80 5G modem satellite

Mobile World Congress 2024 (MWC 2024) has just started, and Qualcomm had three major announcements with the introduction of the Snapdragon X80 5G modem with NB-NTN satellite connectivity, the Qualcomm AI Hub with over 75 AI models optimized for Snapdragon processors, and the FastConnect 7900 WiFi 7, Bluetooth, and Ultra Wideband (UWB) chip. Snapdragon X80 5G modem-RF system Snapdragon X80 5G modem specifications: Peak download speed – 10 Gbps Peak upload speed – 3.5 Gbps Cellular Modem-RF Specs – 10CC aggregation in mmWave, 5CC aggregation in 5G sub-6GHz Cellular Technology 5G NR, sub-6 GHz Dynamic Spectrum Sharing (DSS) LTE, WCDMA, LAA, TD-SCDMA, GSM/EDGE, CBRS mmWave, sub-6 carrier aggregation (FDD-FDD) sub-6 carrier aggregation (TDD-TDD), 5G FDD, 5G TDD, sub-6 carrier aggregation (FDD-TDD) 5G SA (standalone), 5G NSA (non-standalone) F + F ULCA, FDD UL MIMO, Switched Uplink, 3GPP R18-ready (5G Advanced-ready), 6x CA sub-6 downlink carrier aggregation, 1024-QAM sub-6, FR1 + […]

Posted on by Debashis Das - No Comments on BrainChip’s Neuromorphic Akida Edge AI Box is now available for pre-orders at $799

BrainChip’s Neuromorphic Akida Edge AI Box is now available for pre-orders at $799

BrainChip Neuromorphic Akida Edge AI Box

BrainChip has recently opened preorders for their Akida Edge AI Box, built in partnership with VVDN Technologies. This box features an NXP i.MX 8M Plus SoC and two Akida AKD1000 neuromorphic processors for low-latency, high-throughput AI processing at the edge. The system features USB 3.0 and micro-USB ports, HDMI, 4GB LPDDR4 memory, 32GB eMMC with up to 1TB micro-SDXC expansion, dual-band Wi-Fi, and two gigabit Ethernet ports for external camera connections, all within a compact, passively-cooled chassis, powered by 12V DC. BrainChip Akida Edge AI Box Specifications: Host CPU – NXP i.MX 8M Plus Quad SOC with 64-bit Arm Cortex-A53 processor running at up to 1.8GHz AI/ML Accelerator – Dual Brainchip AKD1000 (Akida Chip) over PCIe for efficient AI processing Memory – 4GB LPDDR4 Storage 32GB eMMC flash MicroSD card slot for additional storage options Display Output – HDMI output supporting up to 3840 x 2160p30 resolutions with a pixel clock […]

Posted on by Jean-Luc Aufranc (CNXSoft) - 3 Comments on Doly – A cute little autonomous AI-powered robot based on Raspberry Pi CM4 module (Crowdfunding)

Doly – A cute little autonomous AI-powered robot based on Raspberry Pi CM4 module (Crowdfunding)

Doly Raspberry Pi CM4 robot

Limitbit Doly is a cute little autonomous robot with two continuous tracks, two small arms controlled by servos, two round color displays acting as the eyes, and various sensors, all controlled by a Raspberry Pi CM4 system-on-module. The robot can be used for STEM (Science, Technology, Engineering, and Mathematics) education or as a developer platform. AI workloads can also run on the Raspberry Pi CM4 module taking sensors, camera, and microphone inputs, with the robot interacting with the user through the built-in stereo speaker and two eyes. In practice, that means Doly supports features such as face recognition and smart audio with the robot capable of recognizing its owner and responding to voice commands. Doly specifications: System-on-Module – Raspberry Pi CM4 Lite model CM4101000 (1GB RAM, Wireless) by default, but also supports other CM4/CM Lite modules with wireless Storage – MicroSD card slot Display – 2x high-resolution color displays (the […]

Posted on by Jean-Luc Aufranc (CNXSoft) - 2 Comments on Luxonis OAK Thermal – A PoE thermal camera with Myriad X AI accelerator, waterproof M12 and M8 connectors

Luxonis OAK Thermal – A PoE thermal camera with Myriad X AI accelerator, waterproof M12 and M8 connectors

Luxonis OAK Thermal camera

Luxonis has announced its first thermal camera with the OAK Thermal (OAK-T) based on the company’s OAK-SoM Pro AI module featuring an Intel Movidius Myriad X, and two waterproof ports with an M12 PoE/Ethernet connector and an M8 auxiliary connector. Luxonis has been making AI cameras based on Myriad X AI accelerator and its Depth AI solution at least since 2019, and its module is also found in third-party cameras as we’ve recently found out with the Arducam PiNSIGHT AI camera. But they had never made a thermal model, and following customers’ requests to fuse thermal and RGB data, they’ve now developed the OAK Thermal, or OAK-T for shorts, that is suitable for detecting leaks and fires or more accurately detect humans & animals than traditional vision-only based cameras. OAK Thermal camera specifications: System-on-Module – Luxonis OAK-SoM Pro with AI accelerator – Intel Movidius Myriad X AI vision processing unit […]

Posted on by Jean-Luc Aufranc (CNXSoft) - 21 Comments on Duo 256M is a compact SBC based on SG2002 multi-architecture SoC

Duo 256M is a compact SBC based on SG2002 multi-architecture SoC

Duo 256M SBC

Duo 256M is a small board powered by SOPHGO SG2002 multi-architecture Arm/RISC-V/8051 SoC with 256MB of on-chip RAM and a 1 TOPS NPU, a microSD card for storage, a camera connector, a USB-C port for power and programming, and two headers for GPIO expansion. We covered the SOPHGO SG2002 (and SG2000)  Arm+RISC-V+8051 AI SoC earlier this month saying a couple of boards were expected very soon. We’ve already covered Sipeed LicheeRV Nano with optional Ethernet or WiFi 6, and now we’ll look at the Duo 256M designed by Milk-V Technology in more detail since it’s available now. Duo 256M specifications: SoC – SOPHGO SG2002 Main core – 1GHz 64-bit RISC-V C906 or Arm Cortex-A53 core (selectable) Minor core – 700MHz 64-bit RISC-V C906 core Low-power core – 25 to 300MHz 8051 MCU core NPU – 1 TOPS INT8, supports BF16 Integrated 256MB DDR3 (SiP) Storage MicroSD card slot 32Gbit NAND […]

Posted on by Jean-Luc Aufranc (CNXSoft) - No Comments on Femto Mega 3D depth camera review – OrbbecViewer and Orbbec SDK in Ubuntu 22.04, body tracking in Windows 11

Femto Mega 3D depth camera review – OrbbecViewer and Orbbec SDK in Ubuntu 22.04, body tracking in Windows 11

Orbbec Femto Mega Review

We had a quick look at the Orbbec Femto Mega 3D depth and 4K RGB camera at the end of last year with a look at the hardware and a quick try with the OrbbecViewer program in Ubuntu 22.04. I’ve now had the time to test the OrbbecViewer in more detail, check out the Orbbec SDK in Linux and various samples, and finally test the Femto Mega with a body tracking application using Unity in Windows 11. A closer look at OrbbecViewer program and settings As noted in the first part of the review, the OrbbecViewer program provides color, depth, and IR views for the camera. Let’s go into details for each. The Color mode would be the same as your standard USB color camera. The Femto Mega supports resolution from 1280×720 at 30 fps to 3840×2160 at 25 fps using MJPG, H.264, H.265, or RGB (converted from MJPG). We […]

Posted on by Jean-Luc Aufranc (CNXSoft) - 2 Comments on Arducam PiNSIGHT – A 4 TOPS AI camera board for the Raspberry Pi 5

Arducam PiNSIGHT – A 4 TOPS AI camera board for the Raspberry Pi 5

Arducam PiNSIGHT Raspberry Pi 5 AI Camera module

Arducam PiNSIGHT is an AI camera board designed for the Raspberry Pi 5 equipped with a 12.3MP auto-focus module and an Intel Movidius Myriad X-powered SoM delivering up to 4 TOPS and supporting Intel OpenVINO deep learning models. The PiNSIGHT “AI Mate” is mounted underneath the Raspberry Pi 5 to which it is connected through a 15-cm USB-A to USB-C cable. The case is made of metal and acts as a heatsink, and Arducam told CNX Software it dissipates heat enough to cool down the Raspberry Pi 5, and they don’t think the active cooler is needed anymore. In other words, it also acts as a fanless enclosure albeit it’s not in contact with the CPU, so whether it’s enough probably depends on your use case… Arducam PiNSIGHT “AI Mate” specifications: AI accelerator – Luxonis OAK-SoM based on Intel Movidius Myriad X vision processing unit delivering up to 4 TOPS […]

UP 7000 x86 SBC