A novel deep neural architecture for efficient and scalable multidomain image classification

S M Nuruzzaman Nobel; Md All Moon Tasir; Humaira Noor; Muhammad Mostafa Monowar; Md Abdul Hamid; Md Shohel Sayeed; Md Rajibul Islam; M F Mridha; Nilanjan Dey

doi:10.1038/s41598-025-10517-w

A novel deep neural architecture for efficient and scalable multidomain image classification

Sci Rep. 2025 Sep 26;15(1):33050. doi: 10.1038/s41598-025-10517-w.

Authors

S M Nuruzzaman Nobel^{1

2}, Md All Moon Tasir^{3

2}, Humaira Noor^{4

2}, Muhammad Mostafa Monowar⁵, Md Abdul Hamid⁵, Md Shohel Sayeed⁶, Md Rajibul Islam⁷, M F Mridha⁸, Nilanjan Dey⁹

Affiliations

¹ School of Engineering, Electrical and Robotics Engineering, Monash University Malaysia, Bandar Sunway, 47500, Subang Jaya, Malaysia.
² RABBIT AI, Selangor, Malaysia.
³ Faculty of Information Science and Technology, Universiti Kebangsaan Malaysia, Bangi, Malaysia.
⁴ Computer Science and Engineering, Michigan State University, Michigan, USA.
⁵ Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah, Saudi Arabia.
⁶ Centre for Intelligent Cloud Computing (CICC), CoE for Advanced Cloud, Faculty of Information Science & Technology, Multimedia University, Melaka, Malaysia. shohel.sayeed@mmu.edu.my.
⁷ Department of Electrical and Electronic Engineering, The Hong Kong Polytechnic University, Hong Kong SAR, China.
⁸ Department of Computer Science, American International University Bangladesh, Dhaka, Bangladesh.
⁹ Techno International New Town, Kolkata, India.

Abstract

Deep learning has significantly advanced the field of computer vision; however, developing models that generalize effectively across diverse image domains remains a major research challenge. In this study, we introduce DeepFreqNet, a novel deep neural architecture specifically designed for high-performance multi-domain image classification. The innovative aspect of DeepFreqNet lies in its combination of three powerful components: multi-scale feature extraction for capturing patterns at different resolutions, depthwise separable convolutions for enhanced computational efficiency, and residual connections to maintain gradient flow and accelerate convergence. This hybrid design improves the architecture's ability to learn discriminative features and ensures scalability across domains with varying data complexities. Unlike traditional transfer learning models, DeepFreqNet adapts seamlessly to diverse datasets without requiring extensive reconfiguration. Experimental results from nine benchmark datasets, including MRI tumor classification, blood cell classification, and sign language recognition, demonstrate superior performance, achieving classification accuracies between 98.96% and 99.97%. These results highlight the effectiveness and versatility of DeepFreqNet, showcasing a significant improvement over existing state-of-the-art methods and establishing it as a robust solution for real-world image classification challenges.

Keywords: Blood cells; Computer vision; Deep learning; Hand sign; MRI tumor classification; Transfer learning; Vision transformer.