Enter fullscreen mode

Edge AI is often seen as a field reserved for powerful single-board computers, but what if you could run machine learning logic directly on a standard Arduino?

In this article, we will explore NocML, an efficient machine learning library specifically designed for resource-constrained microcontrollers.

What is NocML?

NocML is a lightweight C++ library built to bridge the gap between complex ML logic and the limited processing power of microcontrollers like the ESP32, Arduino Uno, or Nano. Inspired by the Scikit-Learn API, it offers a familiar workflow for developers coming from a Python background.

Key Features:

Low Memory Footprint: Optimized to run within the tight SRAM limits of common microcontrollers.
Data Preprocessing: Includes tools like MinMaxScaler for data normalization on-device.
Versatile Algorithms: Supports Classification (KNN, Naive Bayes), Clustering (K-Means), and Regression (Linear Regression).
Zero Latency: Perform inference locally on the device, ensuring privacy and real-time response without cloud dependency.

Use Case: K-Nearest Neighbors (KNN) Classification

One of the strongest features of NocML is its ability to perform sensor classification directly on the "edge." Imagine building a device that classifies activity types based on accelerometer data.

Here is a practical example of how to implement a KNN algorithm using NocML:

#include <NocML.h>

// Define training data (Features)
float X_train = {
  {1.0, 2.0}, // Category A
  {1.5, 1.8}, // Category A
  {5.0, 8.0}, // Category B
  {6.0, 7.0}  // Category B
};

// Labels for training data
int y_train = {0, 0, 1, 1};

// Initialize KNN with k=3
KNN knn(3);

void setup() {
  Serial.begin(115200);

  // Local training (Fit)
  knn.fit((float*)X_train, y_train, 4, 2);
  Serial.println("KNN Model Ready!");
}

void loop() {
  // New sensor data to classify
  float input = {1.2, 2.1};

  // Perform prediction
  int prediction = knn.predict(input);

  Serial.print("Classification Result: ");
  Serial.println(prediction == 0 ? "Category A" : "Category B");

  delay(2000);
}

Why NocML?

Smart IoT: Transform passive sensors into intelligent nodes that make decisions without a server.
Predictive Maintenance: Detect anomalous vibration patterns in industrial motors before failure occurs.
Human-Machine Interaction: Recognize gestures or simple audio patterns in real-time.

Getting Started

You can easily integrate NocML into your project via the Arduino Library Manager or by visiting the official repositories.

Installation

Open your Arduino IDE.
Go to Sketch -> Include Library -> Manage Libraries...
Search for "NocML" and click install.

Alternatively, explore the source code and documentation:

GitHub Repository: Nocturnailed-Community/NocML
Arduino Library Registry: NocML on Arduino Libraries

Conclusion

NocML is part of the TinyML movement, making artificial intelligence accessible on devices costing only a few dollars. By bringing logic closer to the data source, we create faster, more reliable, and smarter IoT systems.

Are you working on an Edge AI project? Give NocML a try and share your results!

#Arduino #MachineLearning #TinyML #IoT #OpenSource #NocLab #ArtificialIntelligence

Bringing Intelligence to the Edge: Introduction to NocML for Arduino

Introduction