N-EIoU-YOLOv9: A Signal-Aware Bounding Box Regression Loss for Lightweight Mobile Detection of Rice Leaf Diseases

TL;DR

This work tackles the challenge of detecting small, low-contrast rice leaf diseases on edge devices by proposing N-EIoU, a signal-aware bounding box regression loss that blends non-monotonic gradient focusing (N-IoU) with geometric decoupling (EIoU). Integrated into the lightweight YOLOv9t framework, the approach leverages PGI and Float16 quantization to enable robust edge inference. Experimental results on a field dataset (5,908 images across four diseases plus Healthy) show a 4.3-point improvement in mAP@50 over CIoU (to 90.3%) and significant gains in tight localization (mAP@50-95 up to 48.9%), with notable improvements for hard-sample classes like Brown Spot and Leaf Blast. The model achieves practical mobile performance (≈156 ms per frame on Android), demonstrating the viability of accurate, offline disease monitoring in agricultural settings and laying groundwork for further dataset expansion and model optimization.

Abstract

In this work, we propose N EIoU YOLOv9, a lightweight detection framework based on a signal aware bounding box regression loss derived from non monotonic gradient focusing and geometric decoupling principles, referred to as N EIoU (Non monotonic Efficient Intersection over Union). The proposed loss reshapes localization gradients by combining non monotonic focusing with decoupled width and height optimization, thereby enhancing weak regression signals for hard samples with low overlap while reducing gradient interference. This design is particularly effective for small and low contrast targets commonly observed in agricultural disease imagery. The proposed N EIoU loss is integrated into a lightweight YOLOv9t architecture and evaluated on a self collected field dataset comprising 5908 rice leaf images across four disease categories and healthy leaves. Experimental results demonstrate consistent performance gains over the standard CIoU loss, achieving a mean Average Precision of 90.3 percent, corresponding to a 4.3 percent improvement over the baseline, with improved localization accuracy under stricter evaluation criteria. For practical validation, the optimized model is deployed on an Android device using TensorFlow Lite with Float16 quantization, achieving an average inference time of 156 milliseconds per frame while maintaining accuracy. These results confirm that the proposed approach effectively balances accuracy, optimization stability, and computational efficiency for edge based agricultural monitoring systems.

Figures (8)

• Figure 1: Graphical Abstract
• Figure 2: Common rice leaf diseases (a, b, c, d) in Vietnam.
• Figure 3: YOLOv9 structure
• Figure 4: Gradient response curves for IoU-based losses
• Figure 5: Gradient magnitude comparison: N-EIoU generates larger gradients in the low-IoU interval ($<0.4$), amplifying signals for hard samples.