CR-YOLOv7: A Lightweight Agricultural Pest Larvae Detection Algorithm Based on Channel Recalibration

• Xiaoyong Liu,
• Yimin Zhu,
• Pei Lu,
• Jinsheng Yi,
• Xi Lu,

Vol. 19, No. 11, November 30, 2025
10.3837/tiis.2025.11.012, Download Paper (Free):

• YOLOv7
• Channel Recalibration
• MPDIoU
• VoVGSCSP
• PyQt5

Abstract

Common agricultural pest larvae, such as Fall Armyworm, usually feed on the stems and leaves of gramineous crops, causing serious economic losses. Therefore, accurate and timely identification and detection of crop pest larvae is the key to pest control. The current detection models have problems such as complex structure and low accuracy. In order to effectively cope with the pest larvae detection task, this study proposes a lightweight detection method CR-YOLOv7 with channel recalibration based on YOLOv7. Specifically, a lightweight network CRNet is proposed as the backbone network of YOLOv7, which recalibrates the feature channels through shuffling operations and attention mechanisms, and applies new weights to the channels step by step, thereby enhancing the ability to capture global features. Then, by integrating the VoVGSCSP module in the head of YOLOv7, the detection accuracy and speed are significantly improved while reducing the computational load. Meanwhile, the MPDIoU loss function is used to improve the convergence speed and regression accuracy. In addition, a larvae dataset (APL8) of eight pests including Fall Armyworm was constructed, and the agricultural pest detection system was implemented using PyQt5, making preliminary explorations for further deployment of related equipment. Experimental results show that compared with YOLOv7, CR-YOLOv7 achieves 94.85% mAP@0.5 on the APL8 dataset, and the FPS is improved by 15.79%, and it also performed well on the IP102 dataset. This study not only provides a feasible solution for pest larvae detection, but also has practical significance.

Statistics

Cite this article