WSEAS Transactions on Signal Processing
Print ISSN: 1790-5052, E-ISSN: 2224-3488
Volume 22, 2026
Spectra-MobileNet: A Frequency-Aware Ordinal Framework for Fine-Grained Prawn Freshness Assessment
Authors: , , , , , , , ,
Search Articles
Abstract: The freshness assessment of prawns has become imperative in the aquaculture industry. Manual methods
that were previously used have been found to be highly inconsistent and expensive too. This implies that there
is need for a reliable and efficient technique like the automated image classification, which will carry out the process
quickly. Researches related to shrimp quality assessment have largely relied on datasets that are inadequate
and only focused on visual features that cannot adequately distinguish freshness differences. To overcome the
limitations, the developed Spectra-MobileNet technique has been able to apply lightweight design and frequency
domain feature extraction to achieve effective freshness recognition. In contrast to conventional Convolutional
Neural Networks, which focus solely on spatial attributes, the technique relies on frequency domain images with
MobileNetV2 architecture. The applicability of the method is verified via an experiment using the prawn images
dataset.
Keywords:
Prawn Quality Classification, MobileNetV2, Convolutional Block Attention Module (CBAM), Deep Learning, Attention Mechanism, Aquaculture, Image Classification, Computer Vision
Pages: 193-205
DOI: 10.37394/232014.2026.22.17