Skip to content

afondiel/Edge-AI-Platforms

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 

Repository files navigation

Edge AI Platforms

Overview

A curated list of Edge AI hardware platforms that support AI and machine learning developments for real-world applications.

Table of Contents

Edge Device Platforms

1. System on Chips (SoCs):

  • NVIDIA Jetson Series:
    • Jetson Orin Nano Developer Kit: Compact AI computer with 67 TOPS for robotics and AI applications.
    • Jetson Xavier NX: Up to 21 TOPS for edge AI applications.
    • Jetson Nano: Entry-level AI device for robotics and vision tasks.
  • Qualcomm Snapdragon Series:
    • Snapdragon 8cx Gen 3: AI processing for mobile and embedded devices.
    • Snapdragon 888 AI Accelerator: Built for high-performance on-device AI.
  • Apple Silicon:
    • M1/M2 Series: Integrated AI capabilities for edge computing.
  • Google Coral Edge TPU: Specialized for AI inference at the edge.
  • Hailo-8 AI Processor: High-performance deep learning on edge devices.
  • Rockchip RK3399Pro: Combines CPU, GPU, and AI processing.
  • HiSilicon Ascend Series: AI-focused SoCs for industrial and mobile applications.

2. Microcontrollers (MCUs):

  • STMicroelectronics STM32N6 Series: AI microcontrollers for lightweight edge tasks (image/audio).
  • Texas Instruments SimpleLink™ MCU Series: AI and ML-ready for IoT and edge computing.
  • Espressif ESP32 Series: Compact AI-enabled microcontroller for lightweight applications.
  • NXP i.MX RT Series: Supports AI tasks on ultra-low-power microcontrollers.
  • Renesas RA MCU Series: Optimized for TinyML workloads in IoT and edge systems.

3. Field-Programmable Gate Arrays (FPGAs):

  • Xilinx Alveo Series: High-performance AI workloads at the edge.
  • Intel Stratix 10 Series: Designed for AI inference in resource-intensive tasks.
  • EdgeLLM Accelerator: Heterogeneous CPU-FPGA for large language models at the edge.

4. Edge AI Boxes and Gateways:

  • Qualcomm Edge AI Box: Industrial edge AI device for vision and sensor applications.
  • NVIDIA Jetson-based Edge AI Boxes: Pre-configured AI deployment systems.
  • Advantech Edge AI Box PCs: Designed for industrial and smart city applications.
  • Intel NUC Kits: AI-ready edge computing boxes.

5. Mobile and Embedded Devices:

  • AI-enabled Smartphones:
    • Apple's iPhones with Neural Engine.
    • Android phones with AI-dedicated processors (e.g., Google Tensor).
  • Raspberry Pi with AI Accelerators:
    • Paired with Google Coral USB Accelerator or Intel Neural Compute Stick.
  • BeagleBone AI-64: Embedded system for AI development.
  • Khadas VIM3: Linux-powered AI board for edge applications.
  • Jetson Nano Embedded Systems: Customized AI systems for robotics and IoT.

6. AI Development Platforms:

  • Edge Impulse: AI development on microcontrollers and SoCs.
  • AWS SageMaker Edge: Optimized ML models for edge deployment.
  • TensorFlow Lite: Framework for deploying ML models on edge devices.
  • PyTorch Mobile: For edge AI model development and deployment.

7. Specialized Edge Devices:

  • Google Nest Devices: Smart home devices with AI integration (e.g., Google Nest Cam).
  • Amazon Echo Devices: AI-powered assistants with on-device processing.
  • Microsoft Azure Percept: AI-powered platform for edge solutions.
  • AI Surveillance Cameras: Devices like Hikvision AI-powered security systems.

8. Industrial and Custom Edge Devices:

  • Bosch IoT Suite Devices: AI-ready industrial IoT systems.
  • Siemens MindSphere Gateways: Integrated edge AI for industrial automation.
  • Intel Movidius Myriad X: Vision processing units for industrial AI applications.
  • Advantech Embedded AI Systems: Specialized for factory automation and logistics.

9. Robotics-focused Edge Devices:

  • Unitree Robotics A1: Robot with integrated edge AI processing.
  • Boston Dynamics Spot: AI-powered robot with edge computing capabilities.
  • Open Robotics TurtleBot3: Affordable robot with AI-enabled edge processing.

Back to Table of Contents

Industry Applications

1. Industrial Automation

Edge devices optimized for manufacturing, logistics, and factory automation:

  • NVIDIA Jetson AGX Xavier: Robotic arms, smart manufacturing.
  • Bosch IoT Suite Devices: Real-time monitoring, predictive maintenance.
  • Intel Movidius Myriad X: AI vision for industrial automation.
  • Advantech Embedded AI Systems: Factory optimization, process control.
  • Xilinx Alveo Series: High-speed, real-time industrial AI.

2. Healthcare and Medical Devices

Edge devices used for diagnostics, medical imaging, and remote healthcare:

  • Qualcomm Snapdragon 8cx Gen 3: AI in portable medical devices.
  • Google Coral Edge TPU: Real-time diagnostics and analysis.
  • Jetson Orin Nano Developer Kit: AI in point-of-care devices.
  • Siemens Healthineers Edge Systems: Imaging and diagnostic machines.
  • STMicroelectronics STM32N6 Series: TinyML in wearable health trackers.

3. Automotive and Transportation

Edge solutions for ADAS, autonomous driving, and fleet management:

  • NVIDIA Jetson Xavier NX: Autonomous vehicles, ADAS.
  • Qualcomm Snapdragon Ride Platform: Automotive AI workloads.
  • Hailo-8 AI Processor: Object detection and path planning in vehicles.
  • Intel NUC Kits: In-car edge computing for infotainment.
  • Renesas R-Car Series: Traffic monitoring, vehicle-to-everything (V2X).

4. Smart Cities and Surveillance

AI devices for urban infrastructure, traffic management, and security:

  • Google Nest Cam: Real-time AI-enabled surveillance.
  • Hikvision AI Surveillance Cameras: Smart security systems.
  • Advantech Edge AI Boxes: Urban planning and monitoring.
  • Jetson Nano Embedded Systems: Pedestrian and vehicle tracking.
  • Amazon Echo Devices: Smart city IoT integration.

5. Retail and E-commerce

Devices for inventory management, personalized shopping experiences, and checkout-free stores:

  • NVIDIA Jetson Nano: AI-powered checkout systems.
  • Intel Movidius Myriad X: Shelf scanning and customer behavior analysis.
  • Advantech Edge AI Box PCs: Automated inventory tracking.
  • Khadas VIM3: AI in cashier-less stores.
  • STMicroelectronics STM32 AI Series: Embedded AI for smart kiosks.

6. Robotics

Edge platforms enabling autonomous operation, AI, and smart behaviors in robotics:

  • NVIDIA Jetson Orin Nano Developer Kit: Autonomous robots.
  • Boston Dynamics Spot: Industrial and field robotics with edge AI capabilities.
  • NVIDIA Isaac Robot Platform: AI-driven simulation and deployment for robotics.
  • Clearpath Robotics Jackal UGV: Unmanned ground vehicle for autonomous navigation.
  • Fetch Robotics Freight 500: AI-driven logistics and warehouse automation.
  • KUKA iiwa Robot: AI-enhanced industrial collaborative robots.
  • Unitree Robotics A1: Quadruped robot with AI for movement and sensing.
  • Open Robotics TurtleBot3: Research and educational robots.
  • BeagleBone AI-64: Robotics prototyping with on-device AI.

7. Agriculture

AI-enabled devices for precision farming, crop monitoring, and livestock management:

  • Jetson Xavier NX: AI in drones for agriculture.
  • Google Coral Edge TPU: Crop health assessment via AI.
  • Raspberry Pi with AI Accelerators: Soil and weather monitoring.
  • Hailo-8 AI Processor: Smart irrigation and yield prediction.
  • Advantech Embedded Systems: Automated harvesting systems.

8. Consumer Electronics and Wearables

AI devices for personal use and smart home integration:

  • Apple Neural Engine (M1/M2): AI in iPhones, iPads, and wearables.
  • Amazon Echo Devices: Smart speakers with on-device AI.
  • Espressif ESP32 Series: Home automation with AI.
  • Google Nest Devices: Smart home systems.
  • STMicroelectronics AI-ready MCUs: Lightweight AI wearables.

9. Edge AI for IoT

Devices driving IoT applications across sectors:

  • Qualcomm Edge AI Boxes: IoT gateways for industrial settings.
  • AWS SageMaker Edge: Cloud-enabled IoT applications.
  • NXP i.MX RT Series: Smart sensors and IoT edge processing.
  • TensorFlow Lite-enabled Devices: ML for IoT sensors and cameras.
  • PyTorch Mobile on Raspberry Pi: IoT prototyping with AI models.

Back to Table of Contents

Edge Devices Categorized by Computing Performance

1. Light Performance Applications

Devices in this category are suitable for basic tasks, low-power applications, or those with simpler AI/ML models that do not require heavy computation.

Device Description CPU Type Memory Storage Energy Inference Speed Latency GPU Support Throughput Use Cases
Raspberry Pi 4 (with AI accelerator) A versatile single-board computer with AI accelerator support for light ML applications. Low-power processors (ARM Cortex-A) < 1 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low Simple sensors, basic monitoring
Google Coral Dev Board (Edge TPU) Edge device with Edge TPU for accelerating ML tasks like image recognition at low power. Low-power processors (ARM Cortex-A) 1 GB - 4 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low IoT, basic AI
BeagleBone AI-64 AI-capable development board offering basic edge AI capabilities. Low-power processors (ARM Cortex-A) 1 GB - 4 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low Simple robotics, monitoring systems
ESP32 (with AI capabilities) Low-power microcontroller with built-in Wi-Fi and Bluetooth, capable of running basic AI algorithms. Low-power processors (low-end RISC) < 1 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low Simple IoT devices, wearables
Arduino Portenta H7 High-performance board for edge computing tasks, suitable for moderate AI workloads with low power requirements. Low-power processors (ARM Cortex-A) 1 GB - 4 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low Robotics, smart home
NVIDIA Jetson Nano Entry-level edge AI platform with GPU support for basic AI tasks. Low-power processors (ARM Cortex-A) 1 GB - 4 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low Basic AI, robotics
Unitree Robotics A1 Lightweight robot designed for basic tasks and low-end AI applications. Low-power processors (ARM Cortex-A) < 1 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low Basic robot navigation
Open Robotics TurtleBot3 Simple, open-source robot with basic sensors and light AI capabilities. Low-power processors (ARM Cortex-A) < 1 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low Educational robotics
Qualcomm Snapdragon 410e Embedded AI solution targeting low-power, mobile applications. Low-power processors (ARM Cortex-A) 1 GB - 4 GB Flash storage Low power (1 - 5 W) 500 ms - 2 s High latency Basic GPU Low Simple embedded AI systems

2. Mid Performance Applications

These devices can handle more computationally intensive tasks and are suited for real-time AI inference, edge computing, and moderate machine learning tasks.

Device Description CPU Type Memory Storage Energy Inference Speed Latency GPU Support Throughput Use Cases
NVIDIA Jetson Xavier NX Powerful edge AI device with the ability to handle real-time AI inference. Mid-range processors (ARM Cortex-A) 4 GB - 8 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium Robotics, AI edge inference
NVIDIA Jetson Orin Nano Developer Kit More advanced edge device for running demanding AI models with low power. Mid-range processors (ARM Cortex-A) 4 GB - 8 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium Robotics, AI processing
Google Coral Edge TPU Edge device designed for running fast, efficient ML models with a dedicated TPU for inference. Mid-range processors (ARM Cortex-A) 4 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium AI edge, machine vision
Raspberry Pi 4 (with AI accelerator) Entry-level AI device with sufficient power for moderate AI inference. Mid-range processors (ARM Cortex-A) 4 GB Flash storage Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium AI for IoT systems
Intel NUC (Edge AI capabilities) Compact PC designed to run AI models at the edge with decent processing power. Mid-range processors (x86-based) 4 GB - 8 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium AI applications, robotics
KUKA iiwa Robot Advanced robotic platform with AI for industrial automation tasks. Mid-range processors (Intel Atom) 8 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium Industrial automation, robots
Clearpath Robotics Jackal UGV Autonomous robot designed for outdoor, rugged environments with real-time AI navigation. Mid-range processors (ARM Cortex-A) 4 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium Outdoor robotics, AI navigation
Boston Dynamics Spot Advanced robot designed for various industrial applications with AI-powered perception and autonomy. Mid-range processors (ARM Cortex-A) 8 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium Robotics, navigation, AI tasks
Fetch Robotics Freight 500 Logistics robot using AI for material transport in warehouses. Mid-range processors (ARM Cortex-A) 4 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium Logistics, transport
NVIDIA Jetson TX2 Edge computing platform suitable for AI and robotics applications with moderate processing needs. Mid-range processors (ARM Cortex-A) 8 GB SSD Moderate power (5 - 30 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium AI, robotics, automation
Intel Movidius Neural Compute Stick 2 USB stick with AI acceleration for running models at the edge. Low-power processors < 1 GB Flash storage Low power (1 - 5 W) 50 ms - 500 ms Moderate latency Integrated GPU Medium Edge AI inference

3. High Performance Applications

Devices in this category offer the highest computational power and are suitable for complex AI/ML models, heavy real-time processing, and advanced robotics.

Device Description CPU Type Memory Storage Energy Inference Speed Latency GPU Support Throughput Use Cases
NVIDIA Jetson AGX Orin High-performance edge AI device designed for complex machine learning models and real-time AI tasks. High-end processors (ARM Neoverse) > 8 GB NVMe storage High power (30 W - 150 W) < 50 ms Low latency Dedicated AI accelerators High Autonomous vehicles, real-time video processing, AI for robotics
NVIDIA Jetson Xavier AGX Top-tier edge AI platform for heavy-duty AI tasks and robotics applications. High-end processors (ARM Neoverse) 16 GB SSD High power (30 W - 150 W) < 50 ms Low latency Dedicated GPU High AI in autonomous systems, robotics
Intel Xeon Scalable (with AI accelerator) High-end edge computing platform for advanced AI applications, typically used in data centers. High-end processors (Intel Xeon) > 8 GB SSD High power (30 W - 150 W) < 50 ms Low latency Dedicated AI accelerators High Cloud and edge AI, machine learning, AI applications in robotics
Google TPU (Cloud-based AI inference) Cloud-based solution for running complex machine learning models at scale with low latency. High-end processors (custom TPUs) > 16 GB SSD/NVMe High power (30 W - 150 W) < 50 ms Low latency High throughput High Large-scale AI, autonomous vehicles
IBM Power9 with AI Acceleration High-performance computing platform optimized for AI and deep learning tasks. High-end processors (Power9) > 32 GB SSD/NVMe High power (30 W - 150 W) < 50 ms Low latency High throughput High AI in advanced robotics and industrial applications

Back to Table of Contents

References

About

A curated list of resource-constrained platforms for efficient and local AI developments.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors