Analog Devices' MAX78000 series enables low-power AI (Artificial Intelligence) processing on edge devices, with the ability to handle applications such as machine vision, audio, and facial recognition in real time.
This series can power AI using less than 1/100th the power of other embedded solutions.
The MAX78000 series contains two microcontroller cores (Arm Cortex-M4 and RISC-V) and an accelerator dedicated to Convolutional Neural Networks (CNN). This resolves challenges in AI inference processing at the edge, such as power consumption and latency.

• 

  MAX78000 Internal Block Diagram

• 

  MAX78002 Internal Block Diagram

Dual-core, low-power microcontroller

• Arm Cortex-M4 processor with FPU: up to 120MHz
• 2.5MB Flash, 64KB ROM, and 384KB SRAM
• Optimized performance with a 16KB instruction cache
• Optional Error Correction Code (ECC SEC-DED) for SRAM
• 32-bit RISC-V coprocessor: up to 60MHz
• Up to 60 general-purpose input/output pins
• MIPI Camera Serial Interface 2 (MIPI CSI-2) Controller V2.1: supports two data lanes
• 12-bit parallel camera interface
• I2S controller/target for digital audio interface
• Secure Digital Interface: supports SD 3.0 / SDIO 3.0 / eMMC 4.51

Convolutional Neural Network (CNN) accelerator

• Convolutional Neural Network (CNN) accelerator
• Supports 2 million 8-bit weights (1, 2, 4, and 8-bit weights)
• CNN data memory: 1.3MB
• Programmable input image size: up to 2048 × 2048 pixels
• Programmable network depth: up to 128 layers
• Programmable network channel width per layer: up to 1024 channels
• 1- and 2-dimensional convolution processing
• Able to process VGA images at 30 fps

Security and Integrity

• Secure boot available
• AES 128/192/256 hardware acceleration engine
• True Random Number Generator (TRNG) seed generator

A Gesture Solution Using MAX78000

• Deep learning-based gesture/palm shape determination
  This configuration achieves high-speed processing using MAX78000's AI accelerator for determination.

Reference

• Optical Gesture Solutions | NEXTY System Design

Example Gesture Sensor System

Gesture / Shape Recognition Processing Time Comparison

• Results show over 20× faster processing speed than Cortex M4 cores.

A Gesture Solution Using MAX78000

• Feature extraction method
  Uses MTCNN [1] and FaceNet [2] algorithms for extracting facial features.
• Post-processing
  Estimates the model with the closest features based on features extracted from MTCNN and FaceNet.

Dual-core, low-power microcontroller

• Input device: microphone
• Inference Result: Abnormal sound detection
• System Features
  Performs audio capture and inference every second.
  Provides an edge AI-based small-scale development solution

Examples of abnormal sound detection

• 

  Gunshot / Glass breaking / Siren

• 

  Fall detection

• 

  Mosquito detection

• Input Device: Accelerometer
• Inference Result: Anomaly detection
• System Features
  Realizes an AI model capable of detecting anomalies without requiring error data
  Enables edge AI-based small-scale development solutions
  Makes rapid system development and evaluation possible by using Analog Devices' accelerometer and signal chain.

Examples of abnormal sound detection

• 

  Anomaly monitoring system for industrial motors.