Optimized ReXNet variants with spatial pyramid pooling, CoordAttention, and convolutional block attention module for money plant disease detection
Abstract Money plants, widely appreciated for their ornamental appeal and air-purifying capabilities, are susceptible to several diseases most notably Bacterial Wilt Disease and Manganese Toxicity—that can significantly impact their vitality and aesthetic value. This study explores the effectiveness...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-05-01
|
| Series: | Discover Sustainability |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s43621-025-01241-6 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Money plants, widely appreciated for their ornamental appeal and air-purifying capabilities, are susceptible to several diseases most notably Bacterial Wilt Disease and Manganese Toxicity—that can significantly impact their vitality and aesthetic value. This study explores the effectiveness of optimized ReXNet variants integrated with advanced attention mechanisms, including Coordinate Attention (CoordAttention), spatial pyramid pooling (SPP), and the convolutional block attention module (CBAM), for the precise classification of these diseases alongside healthy specimens. A balanced dataset comprising 15,000 high-resolution images was employed to fine-tune and evaluate the model using precision, recall, F1-score, and overall accuracy as performance metrics. The baseline ReXNet model, after fine-tuning, achieved an accuracy of 88%, serving as a reliable benchmark. The CoordAttention-enhanced ReXNet improved both spatial and channel-wise feature extraction, attaining 96% accuracy by leveraging coordinate-separated pooling to highlight relevant features. The SPP-integrated ReXNet further enhanced performance through multi-scale feature aggregation, reaching 98% accuracy and demonstrating robustness in classifying images with overlapping features. The most effective configuration, ReXNet with CBAM, combined channel and spatial attention mechanisms dynamically, yielding flawless classification with a perfect accuracy of 99%. These results highlight the significant impact of attention-based modules in enhancing feature representation and classification accuracy, offering a lightweight and scalable approach for precision agriculture and ornamental plant disease monitoring. |
|---|---|
| ISSN: | 2662-9984 |