Citation

Lim, J.M and Lim, K.M. and Goh, Pey Yun and Lee, C.P. (2026) Fine-Grained Image Classification with Practical Automated Data Augmentation and ConvNeXt. IAENG International Journal of Computer Science, 53 (9). pp. 3762-3774. ISSN 1819656X

Text 15.pdf - Published Version Restricted to Repository staff only Download (200kB)

Abstract

Fine-grained image classification is inherently challenged by subtle inter-class differences and substantial intra-class variations. To overcome these limitations, this work introduces a high-precision classification framework integrating the ConvNeXt architecture with practical automated data augmentation. By employing RandAugment, the model systematically applies diverse stochastic transformations to improve generalization and suppress overfitting without the need for manual policy design. This augmented data is processed through ConvNeXt, which utilizes depthwise convolutions and modernized normalization to capture highly discriminative feature representations. We evaluate the proposed framework across three benchmark datasets: CUB-200-2011, Stanford Dogs, and Stanford Cars. The method achieves competitive performance of 89.14, 89.28, and 93.33, respectively. Rigorous quantitative analyses via the Davies-Bouldin index and Friedman test, supported by qualitative t-SNE visualizations, demonstrate that the learned embeddings form highly compact and well-separated clusters. Ultimately, this approach proves that coupling automated augmentation with advanced convolutional networks significantly enhances discriminative capability and statistical reliability over conventional CNN and transformer baselines.

Item Type:	Article
Uncontrolled Keywords:	Automated Data Augmentation, ConvNeXT, Deep Learning Models
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Information Science and Technology (FIST)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	02 Oct 2026 06:16
Last Modified:	02 Oct 2026 06:16
URII:	http://shdl.mmu.edu.my/id/eprint/16836

Downloads

Edit (login required)