Augment to Segment: Tackling Pixel-Level Imbalance in Wheat Disease and Pest Segmentation

2025-09-12

AI Summary

In wheat pest and disease imagery, pest-infested regions occupy an extremely low pixel ratio, causing severe class imbalance in semantic segmentation and leading models to overfit dominant classes while neglecting rare pest regions. To address this, we propose a rare-class-oriented data augmentation method: it involves cropping rare-class samples, applying random geometric transformations, non-overlapping pasting, and integrating random projection filtering to achieve natural blending of lesion regions and enhance local feature consistency. Unlike conventional copy-paste approaches, our method eliminates artifacts and overlapping artifacts. Experimental results show that it significantly improves segmentation accuracy for pest regions—achieving up to a 12.6% gain in mean Intersection-over-Union (mIoU)—while maintaining or slightly improving performance on other classes. Extensive evaluation demonstrates the method’s effectiveness and generalizability in fine-grained agricultural disease segmentation tasks.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Search and Optimization: Distributed Search

Application Category

• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Accurate segmentation of foliar diseases and insect damage in wheat is crucial for effective crop management and disease control. However, the insect damage typically occupies only a tiny fraction of annotated pixels. This extreme pixel-level imbalance poses a significant challenge to the segmentation performance, which can result in overfitting to common classes and insufficient learning of rare classes, thereby impairing overall performance. In this paper, we propose a Random Projected Copy-and-Paste (RPCP) augmentation technique to address the pixel imbalance problem. Specifically, we extract rare insect-damage patches from annotated training images and apply random geometric transformations to simulate variations. The transformed patches are then pasted in appropriate regions while avoiding overlaps with lesions or existing damaged regions. In addition, we apply a random projection filter to the pasted regions, refining local features and ensuring a natural blend with the new background. Experiments show that our method substantially improves segmentation performance on the insect damage class, while maintaining or even slightly enhancing accuracy on other categories. Our results highlight the effectiveness of targeted augmentation in mitigating extreme pixel imbalance, offering a straightforward yet effective solution for agricultural segmentation problems.

Problem

• Addresses extreme pixel-level imbalance in segmentation
• Improves rare insect-damage class segmentation performance
• Mitigates overfitting to common classes in agricultural imagery

Innovation

• Random Projected Copy-and-Paste augmentation technique
• Extracts rare patches and applies geometric transformations
• Uses random projection filter for natural blending