Enhancing Rice Bacterial Leaf Blight Detection Accuracy Through RCAMNET

View at publisher Summarise with VIRAT
Authors: Kodeeswari K, Suganya S, S. Sasirekha, M. Jaithoon Bibi Kalaivani, K. Hazeena, S. S. Ananthan
Cited by: 0
References: 15

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Early and reliable detection of Bacterial Leaf Blight (BLB) in rice is essential for disease control and food safety security. Terminal diagnosis at field level, however, remains challenging due to different light conditions, complex backgrounds, and shared symptoms with other stresses. This paper presents RCAMNet, a lightweight Residual + Channel Attention Module Network capable of achieving high detection accuracy with low computational cost for latency sensitive applications at the edge. RCAMNet integrates a task driven channel attention module that adaptively highlights or suppresses disease related spectral and textural channels while suppressing background noise, jointly with residual learning for stable training and feature reuse. The network was trained and tested using a stratified sample of$\sim 6,200$rice leaf images (binary and multi class BLB stages) with aggressive photometric and geometric augmentation. Compared with the baseline architectures (CNN, VGG16, ResNet50, EfficientNetB0, and Transformer), RCAMNet achieved better performance (Accuracy 0.96, Precision 0.95, Recall 0.96, F1 score 0.95, AUC 0.97) with compact size ($\sim 4.2 \mathrm{M}$parameters, 8.5 GFLOPs) to be practical for mobile/edge usage at the same time. Ablation studies verify that the residual connections, as well as the channel attention module, all contribute substantially to the gains.

Fields of Study & Subjects

Smart Agriculture and AI Advanced Neural Network Applications Plant Pathogenic Bacteria Studies

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
938–943
Publication date
21 January 2026
Language
EN