Abstract: To address the issues of high computational cost and limited deployability in resource-constrained scenarios for cucumber disease detection， a lightweight model， WDL‑YOLOv11， was proposed. Based on YOLOv11n， a WaveletPool module was introduced to reduce computational cost while preserving fine‑grained features， a DySample upsampling operation was introduced to enhance the representation of small objects， and a Detect_LSCD detection head was constructed to achieve parameter sharing and multi‑scale feature calibration. Experiments were conducted on 5191 images containing five types of cucumber diseases and healthy leaves. The results showed that WDL‑YOLOv11 attained an mAP@0.5 of 88.2%， with only 4.1 MB parameters and 4.6 GFLOPs of computational cost. In comparison with YOLOv11n， the number of parameters and the computational cost were reduced by 21.2% and 28.1%， respectively， while the accuracy was decreased by only 0.5 percentage points. The model was superior to Rtdetr‑l and YOLO series lightweight models in terms of both parameter count and computational cost， and ranked among the top in detection accuracy， and exhibited good architectural generalization and disease spot focusing capability.