Citation

Lew, Kai Liang and Shane, Lazaroo and Toa, Chean Khim and Kurniawan, Tetuko (2026) Scratch Train for Lightweight Models for Face Mask Detection. International Journal on Robotics Automation and Sciences, 8 (1). p. 87. ISSN 2682-860X

Text 5.pdf - Published Version Restricted to Repository staff only Download (544kB)

Abstract

Automated systems for detecting face mask use in public became urgent during the COVID-19 pandemic. Most existing mask detection research finetunes ImageNet pre-trained backbones on relatively small datasets of masks. This approach raises concerns about model performance in situations with limited computational resources or when external pretrained weights are not accessible. Additionally, there is a limited comparative analysis of recent lightweight architectures under consistent training conditions for mask detection tasks. This paper evaluates four stateof-the-art lightweight architectures for binary mask detection, including RepViT, ShuffleNetV2, EdgeNeXt Small, and EfficientFormer. These models were trained from scratch using identical training protocols on two datasets containing 7,553 and 11,792 RGB images, respectively. Performance was assessed using standardised metrics, including accuracy, precision, recall, and F1-score. Results showed that EdgeNeXt Small achieved the highest accuracy with 0.980 on Dataset 1. RepViT achieved the highest accuracy with 0.944 on Dataset 2. ShuffleNetV2 achieved the fastest inference time, with 0.51 milliseconds on Dataset 1 and 1.19 milliseconds on Dataset 2. It was the smallest model with 1.26 million parameters across all models. RepViT and EdgeNeXt Small achieved higher accuracy than ShuffleNetV2 but required greater computational resources. EfficientFormer underperformed across all evaluation metrics. These findings indicate that extremely lightweight CNNs can excel at mask detection when trained from scratch. The scope is limited to binary classification and workstation-level profiling. On-device measurements and multi-seed variation are not included

Item Type:	Article
Uncontrolled Keywords:	Convolutional neural network, deep learning
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	09 Jul 2026 03:27
Last Modified:	09 Jul 2026 03:27
URII:	http://shdl.mmu.edu.my/id/eprint/16332

Downloads

Edit (login required)