How to install emlearn-micropython on an ESP32 microcontroller?

Use upip or mpremote to install pre-built .mpy files for the xtensawin architecture. Follow the getting started guide on ReadTheDocs for detailed steps, including setting up MicroPython and verifying installation with examples like xor_trees.

emlearn-micropython vs TensorFlow Lite Micro for MicroPython?

emlearn-micropython offers scikit-learn-like models with lower overhead and easier MicroPython integration, ideal for simple classification and DSP. TensorFlow Lite Micro supports more complex neural networks but often requires C integration and more resources. Choose emlearn for efficiency and simplicity on microcontrollers.

Can emlearn-micropython handle image classification?

Yes, it supports Convolutional Neural Networks using the TinyMaix library, as shown in the mnist_cnn example. However, it's best for small image sizes and low-complexity tasks due to microcontroller resource limits.

What microcontrollers are compatible with emlearn-micropython?

Primarily tested on ESP32 series (xtensawin port) and x64 for development. It should work on other MicroPython ports with native module support, but compatibility isn't guaranteed, so check the documentation for your specific board.

How to train a model with emlearn-micropython on-device?

Use on-device learning features like KNN or linear regression by loading data into array structures, calling fit methods, and saving models to flash. Refer to the user guide for API details and examples like har_trees.

Is emlearn-micropython suitable for real-time signal processing?

With inference times down to 100 microseconds and DSP features like IIR filters and FFT, it can handle real-time tasks on capable microcontrollers, but performance varies by hardware and model complexity, so test with your specific setup.

Open-Awesome

emlearn-micropython

MITC0.11.1

Machine Learning and Digital Signal Processing library for MicroPython, enabling TinyML on microcontrollers without C code.

Visit Website GitHub

167 stars20 forks0 contributors

What is emlearn-micropython?

emlearn-micropython is a machine learning and digital signal processing library for MicroPython that enables developers to run efficient ML models on microcontrollers. It provides Python modules for classification, regression, clustering, and signal processing tasks without requiring C programming. The library is designed for TinyML applications on resource-constrained embedded devices.

Target Audience

MicroPython developers working on microcontroller projects who need machine learning capabilities, particularly those building IoT devices, embedded systems, or edge AI applications with limited resources.

Value Proposition

Developers choose emlearn-micropython because it provides a scikit-learn-like experience for embedded systems with minimal overhead, offering installable native modules that don't require firmware rebuilds while maintaining high efficiency with inference times under 100 microseconds.

Overview

Machine Learning and Digital Signal Processing for MicroPython

Use Cases

Best For

Running classification models on ESP32 microcontrollers
Implementing signal processing filters for sensor data
Building TinyML applications with MicroPython
Adding machine learning to IoT devices without C programming
Developing edge AI systems with limited RAM/FLASH
Creating human activity recognition systems with accelerometer data

Not Ideal For

Projects requiring complex deep learning models with millions of parameters
Applications needing real-time inference at kHz frequencies or higher
Teams mandated to use C/C++ exclusively for embedded ML due to performance SLAs
Systems without MicroPython support or preferring bare-metal programming environments

Pros & Cons

Pros

High Efficiency

Delivers inference times as low as 100 microseconds with RAM and FLASH usage under 2 kB, making it ideal for resource-constrained microcontrollers as highlighted in the features.

Easy Deployment

Offers installable native MicroPython modules (.mpy files) without requiring firmware rebuilds or flashing, simplifying integration for developers.

Broad ML and DSP Support

Includes RandomForest, KNN, CNN, FFT, IIR filters, and linear regression, covering common TinyML tasks and signal processing needs.

On-Device Learning

Supports on-device learning with KNN and linear regression, enabling adaptive models without cloud dependency, as noted in the features.

Cons

Limited Model Complexity

Admitted in the philosophy as suited for 'low-complexity' tasks, it lacks support for advanced neural networks or large-scale ML, restricting use cases.

Platform-Specific Testing

Primarily tested on x64 and xtensawin (ESP32) ports, so support for other MicroPython platforms may be experimental or incomplete, as indicated in the status.

Sparse Advanced Documentation

With a 'minimally useful' status, documentation focuses on basics and may lack in-depth tutorials for complex scenarios or edge cases.

Frequently Asked Questions

Related Projects

MicroMLP

A micro neural network multilayer perceptron for MicroPython (used on ESP32 and Pycom modules)

Stars196

Forks29

Last commit5 years ago

upython-chat-gpt

ChatGPT for MicroPython

Stars30

Forks3

Last commit3 years ago

mp_esp_dl_models

Micropython binding for the ESP32 DL AI vision models like face detection / recognition, imagenet classifier or pedestrian (human) detection

Stars17

Forks13

Last commit1 year ago

MicroPython-NeuralNetwork

Neural Network for MicroPython

Stars0

Forks0

Last commit

Community-curated · Updated weekly · 100% open source

Found a gem we're missing?

Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.

Submit a project Star on GitHub