EL-YOLO: An efficient and lightweight real-time detection network for rice field pests

Rice field pests significantly impact the yield and quality of rice. The current pest identification methods perform poorly when dealing with complex background interference and diverse pest scale variations, in addition, existing models are often too large to meet the requirements of real-time detection. To address these challenges, this paper proposes a real-time rice field pest detection model based on EL-YOLO. Firstly, a lightweight feature extraction module, EIM (Efficient Interaction Module), was designed. It efficiently fuses features from different branches through element-wise multiplication, enhancing the model's feature representation capability while reducing the computational burden. Additionally, this paper proposes a lightweight multi-scale detection head based on shared convolutional layers, LSCD (Lightweight Shared Convolutional Detection Head), which effectively addresses the challenges posed by variations in pest scales, reduces the model size, and enhances real-time performance. Experimental results show that the proposed method achieves a mAP@50 of 94.8%, which is 1.3%, 4.0%, and 2.0% higher than YOLOv8n, YOLOv7tiny, and YOLOv5n, respectively. The parameter count and computational load are reduced by 37% and 34% compared to the baseline, with an FPS of 263. These results validate the effectiveness of the proposed method.