Transformer-based detector spots camouflaged soybean pests more accurately

Researchers have developed an enhanced RT-DETR-ResNet18 framework that improves the detection of soybean pests in complex field environments, where small target size, strong camouflage and severe foliage occlusion have long frustrated automated monitoring. The work, published as an open access Original Article in Complex & Intelligent Systems, reports that the method achieves 77.4% mAP@0.5 and a 79.9% F1-score across 12 soybean pest categories, outperforming the evaluated YOLOv11n and YOLOv12s baselines under the reported protocol.

The framework incorporates three targeted components. The Global Semantic Enhancement Encoding Module (CFBC2f) integrates multi-scale contextual modeling to improve semantic representation under occlusion. The Local Region Self-Attention AIFI (LRSAAIFI) module emphasizes fine-grained local textures to reduce feature dilution for small targets. Finally, the Multi-source Feature Adaptive Modulation Module (MFM) adaptively reweights features to reduce background interference. Together, these modules address the weaknesses that traditional convolutional neural networks may exhibit under these challenging field conditions.

For the highly camouflaged Nezara viridula adult, AP@0.5 rises from 77.7% to 91.8%, an absolute improvement of 14.1 percentage points over YOLOv12s.

To test whether the architectural modifications transfer beyond the INSECT12C benchmark, the team also evaluated the method on PASCAL VOC and VisDrone. The authors describe these results as supplementary evidence of transferability under the evaluated settings, and explicitly caution that they should not be read as proof of universal cross-domain generalization.

Subject of Research: A transformer-based detection framework for identifying camouflaged soybean pests in complex field environments.

Article Title: Breaking the camouflage: a transformer-based approach for soybean pest detection in complex field environments

Article References: Zhang, D., Liu, X., & Wan, J. (2026). Breaking the camouflage: a transformer-based approach for soybean pest detection in complex field environments. Complex & Intelligent Systems. https://doi.org/10.1007/s40747-026-02552-5

Image Credits: AI Generated

DOI: 10.1007/s40747-026-02552-5

Keywords: soybean pests, RT-DETR, transformer, object detection, camouflage, computer vision, agriculture, deep learning, Nezara viridula, mAP, pest detection, attention mechanism