Pig counting is an essential activity in the administration of pig farming. Currently, manual counting is inefficient, costly, and unsuitable for systematic analysis. However, research on dynamic pig counting encounters challenges, including inadequate detection accuracy stemming from crowding, occlusion, deformation, and low-light conditions. Target tracking issues characterized by poor accuracy, frequent identity confusion, and false positive trajectories ultimately lead to diminished accuracy in the final counting outcomes. Given these existing limitations, this paper proposes an enhanced algorithm based on the YOLOv8n+Deep SORT model. The ELA attention mechanism, GSConv, and VoVGSCSP lightweight convolution modules are introduced in YOLOv8n, which improve detection accuracy and speed for pig target recognition. Additionally, Deep SORT is enhanced by integrating the DenseNet feature extraction network and CIoU matching algorithm, improving the accuracy and stability of target tracking. Experimental results indicate that the improved Deep SORT-P pig tracking algorithm attains MOTA and MOTP values of 89.2% and 90.4%, respectively, reflecting improvements of 4.2% and 1.7%, while IDSW is diminished by 25.5%. Finally, counting experiments were performed on videos of pigs traversing the farm passage using both the original and improved algorithms. The improved YOLOv8n-EGV+Deep SORT-P algorithm achieved a counting accuracy of 92.1%, reflecting a 17.5% improvement over the original algorithm. Meanwhile, the improved algorithm presented in this study successfully attained stable dynamic pig counting in practical environments, offering valuable data and references for research on dynamic pig counting.

Gear surface defects vary in scale and often exhibit subtle local features that are obscured by complex metallic backgrounds. However, achieving high detection accuracy and robustness with computationally lightweight models remains challenging in complex industrial inspection settings. We propose GSDD-Net, a lightweight framework based on the YOLO architecture for detecting gear surface defects. The proposed framework incorporates four task-oriented components: (1) ADown combines average and max pooling for downsampling to reduce model complexity while preserving the edge and texture details of subtle defects. (2) CSP-EDLAN is integrated into the backbone and neck networks. It uses cross-stage partial connections, multi-gradient path aggregation, and DualConv units to improve gradient propagation, cross-channel interactions, and multiscale feature fusion. (3) C2TSSA uses token statistical self-attention to capture global contextual dependencies with linear computational complexity, improving the representation of subtle defects against complex backgrounds. (4) LQEHead uses localization quality estimation to extract stable localization features from predicted bounding-box distributions and classification confidence scores. GSDD-Net achieves an mAP50 of 73.0% and an mAP50–95 of 38.0% on the real-world GID dataset. On a public gear defect dataset, it achieves 97.8% and 69.6%, respectively. Compared with the baseline model, GSDD-Net increases mAP50 and mAP50–95 on the GID dataset by 3.1 and 2.1 percentage points, respectively. GSDD-Net reduces the parameter count from 2.6 million to 1.9 million and the computational cost from 6.3 to 5.9 GFLOPs. It also achieves an inference speed of 105.7 FPS. These results indicate that GSDD-Net balances detection accuracy and computational efficiency under the evaluated conditions, supporting its potential for industrial gear surface defect detection.