Hea Choon Ngo

Work place: Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka, Melaka, Malaysia

E-mail: heachoon@utem.edu.my


Research Interests: Data Structures and Algorithms, Swarm Intelligence, Planning and Scheduling, Artificial Intelligence, Computer systems and computational processes, Medical Informatics, Computational Science and Engineering


HEA CHOON NGO is a senior lecturer at the Department of Intelligent Computing and Analytics, Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka (UTeM). He received his Bachelor’s degree in Computer Science (Software Developmet) from the Universiti Teknikal Malaysia Melaka (UTeM) in 2004, a Master’s degree in Information Technology from University of New South Wales (UNSW, Sydney) in 2007 and a PhD in Computer Science from Universiti Sains Malaysia (USM) in 2016. His research interests involve computational intelligence,data science and analytics, planning and scheduling, optimization, health informatics and intelligent systems. He is currently a faculty member of the Faculty of Information and Communication Technology of the Universiti Teknikal Malaysia Melaka (UTeM). He is also a member of the Computational Intelligence and Technologies Lab under the Centre for Advanced Computing Technology, UTeM.

Author Articles
Combining Multi-Feature Regions for Fine-Grained Image Recognition

By Sun Fayou Hea Choon Ngo Yong Wee Sek

DOI: https://doi.org/10.5815/ijigsp.2022.01.02, Pub. Date: 8 Feb. 2022

Fine-grained visual classification(FGVC) is challenging task duo to the subtle discriminative features.Recently, RA-CNN selects a single feature region of the image, and recursively learns the discriminative features. However, RA-CNN abandons most of feature regions, which is not only the inefficient but aslo ineffective.To address above issues,we design a noval fine-grained visual recognition model MRA-CNN,which associates multi-feature regions.To improve the feature representation,attention blocks are integrated into the backbone to reinforce significant features;To improve the classification accuracy, we design the feature scale dependent(FSD) algorithm to select the optimal outputs as the classifier inputs;To avoid missing features, we adopt the k-means algorithm to select multiple feature regions.We demonstrate the value of MRA-CNN by expensive experiments on three popular fine-grained benchmarks:CUB-200-2011,Cars196 and Aircrafts100 where we achieve state-of-the-art performance.Our codes can be found at https://github.com/dlearing/MRA-CNN.git.

[...] Read more.
Other Articles