Lightweight Quantized Deep Neural Network for Air Quality Classification with Microcontroller Environment Profiling

DOI

10.2991/978-94-6463-926-1_91How to use a DOI?

Keywords

Air Quality Classification; Binary Weight Network; Microcontroller Profiling

Abstract

Air pollution is a significant environmental and health challenge, with impacts comparable to major global health risks such as smoking. Developing low-cost, power-efficient monitoring solutions is essential for broader accessibility and deployment. This study introduces a lightweight quantized deep neural network for air quality classification, optimized for microcontrollers and other resource-constrained devices. The dataset, sourced from five monitoring stations in Jakarta (2021–2023), includes measurements of six key pollutants along with their corresponding air quality labels. A base model with five dense layers was implemented in TensorFlow platform and achieved an accuracy of 0.95 ± 0.01. The model was then quantized using the Binary Weight Network (BWN) approach, where only the kernels were binarized while the first layer remained in full precision to preserve accuracy. Performance evaluation included accuracy, model size, and computational efficiency, with profiling conducted via the Edge Impulse Python SDK to simulate microcontroller environments. The quantized model achieved 0.91 ± 0.04 accuracy, reduced memory usage by approximately 74% (369 B vs. 1.47 KiB), and maintained similar training times. Profiling showed RAM usage of 1,656 B, ROM usage of 12,472 B, and inference times between 1–7 ms depending on the device. The results confirm the feasibility of deploying the proposed model on low-power embedded platforms with minimal performance degradation.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - I Nyoman Kusuma Wardana
AU  - Ida Bagus Irawan Purnama
AU  - Dewa Ayu Indah Cahya Dewi
AU  - I Gusti Agung Made Yoga Mahaputra
AU  - Setio Basuki
AU  - Dyah Kurniawati Agustika
PY  - 2025
DA  - 2025/12/31
TI  - Lightweight Quantized Deep Neural Network for Air Quality Classification with Microcontroller Environment Profiling
BT  - Proceedings of the International Conference on Applied Science and Technology on Engineering Science 2025 (iCAST-ES 2025)
PB  - Atlantis Press
SP  - 814
EP  - 822
SN  - 2352-5401
UR  - https://doi.org/10.2991/978-94-6463-926-1_91
DO  - 10.2991/978-94-6463-926-1_91
ID  - Wardana2025
ER  -