# MAUPITI: On-Device Prototype-Based Learning on a Smart Infrared Sensor

arXiv: https://arxiv.org/abs/2608.07192 · HTML: https://mlsanity.com/paper/2608.07192/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 18 references that are in the map.
- First posted on arXiv 7 Aug 2026.

## Builds on (most cited first, 11 of 11)

- [Prototypical Networks for Few-shot Learning](https://mlsanity.com/paper/1703.05175/) (2017)
- [Optuna: A Next-generation Hyperparameter Optimization Framework](https://mlsanity.com/paper/1907.10902/) (2019)
- [End-to-End Incremental Learning](https://mlsanity.com/paper/1807.09536/) (2018)
- [PACT: Parameterized Clipping Activation for Quantized Neural Networks](https://mlsanity.com/paper/1805.06085/) (2018)
- [Overcoming Catastrophic Forgetting in Incremental Few-Shot Learning by Finding Flat Minima](https://mlsanity.com/paper/2111.01549/) (2021)
- [TinyOL: TinyML with Online-Learning on Microcontrollers](https://mlsanity.com/paper/2103.08295/) (2021)
- [PLiNIO: A User-Friendly Library of Gradient-based Methods for Complexity-aware DNN Optimization](https://mlsanity.com/paper/2307.09488/) (2023)
- [Efficient Deep Learning Models for Privacy-preserving People Counting on Low-resolution Infrared Arrays](https://mlsanity.com/paper/2304.06059/) (2023)
- [Human Activity Recognition with Low-Resolution Infrared Array Sensor Using Semi-supervised Cross-domain Neural Networks for Indoor Environment](https://mlsanity.com/paper/2403.02632/) (2024)
- [Train-On-Request: An On-Device Continual Learning Workflow for Adaptive Real-World Brain Machine Interfaces](https://mlsanity.com/paper/2409.09161/) (2024)
- [HW-SW Optimization of DNNs for Privacy-preserving People Counting on Low-resolution Infrared Arrays](https://mlsanity.com/paper/2402.01226/) (2024)

## Abstract (arXiv)

Low-resolution infrared (IR) array sensors represent an interesting solution for privacy-preserving human sensing in embedded systems. In this letter, we describe a smart multi-pixel IR sensor integrating a 16×16 thermal MOSFET (TMOS) array and a RISC-V microcontroller extended with low-precision SIMD instructions, capable of on-device learning and continual adaptation for pose and gesture recognition tasks under tight memory and power constraints (<32kB on-chip memory, ≈1.5mW). To avoid the memory overheads of backpropagation and replay buffers, we adopt a prototype-based Nearest Class Mean (NCM) classifier in which a simple Convolutional Neural Network (CNN) encoder is trained and quantized offline, while class prototypes are stored and updated on the device in streaming mode. With experiments on two datasets, we show that this approach yields accuracy on par with a conventional classifier, with negligible latency overheads in both the classification and the prototype update (<0.29% considering both phases), effectively enabling online adaptation of the perception framework.