Noodle 🍜≈🧠

Lightweight neural network inference engine with file streaming support

GitHub page: https://auralius.github.io/noodle/

Noodle provides primitive modular functions for convolution layer, dense layer, pooling and activations. It allows streaming the intermediate activations, weights, and biases from/to SD/FFat/SD_MMC filesystems to overcome RAM limitations. During the development, we typically test Noodle with low-tier MCUs, such as: Arduino Uno R3, UNO R4, Mega256, and some ESP32 variants.

Persistent identifier

DOI: 10.5281/zenodo.16239227

Special Notes

• Training is done with Keras with PyTorch back-end.

• model_exporter.py is used to export the weights/biases into files.

• Although we still use the Arduino Framework, development is done wit Visual Code and PlatformIO.

Authors

• Auralius Manurung — Universitas Telkom, Bandung -- auralius.manurung@ieee.org
• Lisa Kristiana — ITENAS, Bandung

Copyright and license

Code released under the MIT License. Docs released under Creative Commons.