Wood surface defect detection is a critical step in wood processing and manufacturing. To address the performance degradation caused by small targets and multi-scale features in wood surface defect detection, a novel deep learning model is proposed in this study, FDD-YOLO, specifically designed for this task. In the feature extraction stage, the C2f module and the funnel attention (FA) mechanisms are integrated into the design of the C2f-FA module to enhance the model’s ability to extract features of wood surface defects of various sizes. Additionally, the Dual Spatial Pyramid Pooling-Fast (DSPPF) module is developed, and the Context Self-attention Module (CSAM) is introduced to address the limitations of traditional max pooling methods, which often overlook global contextual information when extracting local features, thereby improving the detection of small-scale wood defects. In the feature fusion stage, a Dual Cross-scale Weighted Feature-fusion (DCWF) module is proposed to fuse shallow, deep, and cross-scale features through a weighted summation approach, effectively addressing the challenge of scale variation in wood surface defects. Experimental results demonstrate that the proposed FDD-YOLO model significantly improves detection performance, increasing the mAP of the baseline model YOLOv8 from 78% to 82.3%, a substantial enhancement of 4.3 percentage points. Furthermore, FDD-YOLO outperforms other mainstream defect detection models in terms of detection accuracy. The proposed model demonstrates significant potential for industrial applications by improving detection accuracy, enhancing production efficiency, and reducing material waste, thereby advancing quality control in wood processing and manufacturing.

As a vital cash crop, intelligent monitoring of pomegranate growth stages plays a crucial role in improving orchard management efficiency and yield. Current research on pomegranates primarily focuses on the detection and quality classification of diseases and pests. Furthermore, the difficulty in deploying complex models in practical scenarios hinders the widespread adoption of pomegranate monitoring technology. To address these challenges, this paper proposes a lightweight pomegranate growth stage detection model, YOLO11-PGM, based on YOLO11n. The model integrates several innovative designs, including a Multi-Scale Edge Enhancement (MSEE) module to mitigate the effects of leaf and fruit occlusion, a Slim Shared Convolutional Head (SSCH) to resolve feature inconsistency across different scales in the feature pyramid, and a High-level Screening Feature Pyramid Network (HSFPN) to replace the standard neck network and achieve a balance between accuracy and complexity. Experimental results demonstrate that YOLO11-PGM achieves an accuracy of 92.3%, a recall of 86.3%, and an mAP50 of 94.0% with only 1.63 M parameters, 4.8 G FLOPs, and a model size of 3.7 MB. It outperforms YOLOv5s, YOLOv7, YOLOv8s, YOLOv9s, and YOLOv10s. Compared with the baseline YOLO11n, YOLO11-PGM improves the mAP50 by 2.6%, reduces the number of parameters by 36.9%, decreases computational complexity by 23.8%, and shrinks the model size by 32.7%. This model offers an effective solution for intelligent monitoring of pomegranate growth stages and provides valuable theoretical and technical references for orchard yield prediction, growth monitoring, planting management optimization, and the development of automated harvesting systems.

Detecting small unauthorized UAVs against complex backgrounds is challenging: targets can be just a few pixels wide, background clutter is pervasive, and the detection system must run on resource-constrained edge hardware. This paper presents a lightweight detector built on YOLOv11 that jointly addresses background suppression and edge-deployment efficiency. A serial channel-spatial attention module is embedded into the backbone to sharpen target features and suppress interference. DySample preserves edge details during resolution recovery, and the detection head is restructured by adding a high-resolution P2 branch while removing the large-object P5 head, shifting the focus to small targets without inflating the model. The WIoU v3 loss function stabilizes training by automatically down-weighting low-quality samples. On the DUT Anti-UAV dataset, the proposed model obtains 92.2% mAP@0.5 and 61.5% mAP@0.5:0.95 with only 2.11 M parameters, a reduction of 18.2% relative to the YOLOv11n baseline. Its ability to generalize across different scenarios is verified by cross-dataset evaluation on Det-Fly and LRDDv2. Deployed on an NVIDIA Jetson Orin Nano Super with TensorRT FP16 optimization, the system achieves 37 end-to-end FPS. These findings demonstrate that the presented method significantly improves small-target detection in complex backgrounds while maintaining a lightweight design suitable for real-time edge deployment.