Rong Deng

The Larus ridibundus (L. ridibundus), a prominent part of Kunming’s landscape, attracts many tourists and boosts the local tourism industry. Effective population monitoring of this species matters for wetland environment evaluation, biodiversity conservation, and ecological civilization construction. Currently, L. ridibundus population statistics mainly rely on manual methods, which are labor-intensive and inefficient. To overcome these limits, we propose a deep learning (DL) framework. It automatically recognizes and counts L. ridibundus by combining CDSP2-YOLOv8n with ByteTrack, aiming to efficiently monitor their population metrics. Our framework uses the optimized YOLOv8n model to achieve excellent multi-object detection for this species. It also uses ByteTrack to effectively reduce target loss from occlusion or overlap during the birds’ flight, providing a sophisticated DL approach for population monitoring. Experimental results show the modified CDSP2-YOLOv8n model works well on the collected L. ridibundus multi-object detection dataset. Its mAP@0.5, mAP@0.5:0.95, Precision, and Recall reach 0.9705, 0.6557, 0.9685, and 0.9496, respectively. When combined with ByteTrack, the proposed framework achieved a Multiple Object Tracking Accuracy (MOTA) of 89.7% and a Multiple Object Tracking Precision (MOTP) of 83.5%. It also demonstrated superior performance in terms of IDF1, Mostly Tracked (MT), Mostly Lost (ML), and Identity Switches (IDSs). Compared to manual counting, our framework has an average accuracy of 91.58%, greatly enhancing the efficiency and accuracy of L. ridibundus population monitoring. In summary, we successfully achieved the automated recognition and counting of L. ridibundus. The proposed method accurately identifies and consistently tracks individual birds, enabling effective population counting. It provides a novel and comprehensive technical approach for the monitoring and conservation of this species and demonstrates promising potential for practical applications.

Recent developments in computer vision have significantly enhanced the automation and objectivity of sports analytics. This paper proposes a novel deep learning-based framework for estimating football possession directly from broadcast video, eliminating the reliance on manual annotations or event-based data that are often labor-intensive, subjective, and temporally coarse. The framework incorporates two structurally improved object detection models: YOLOv8-P2S3A for football detection and YOLOv8-HWD3A for player detection. These models demonstrate superior accuracy compared to baseline detectors, achieving 79.4% and 71.1% validation average precision, respectively, while maintaining low computational latency. Team identification is accomplished through unsupervised DBSCAN clustering on jersey color features, enabling robust and label-free team assignment across diverse match scenarios. Object trajectories are maintained via the Norfair multi-object tracking algorithm, and a temporally aware refinement module ensures accurate estimation of ball possession durations. Extensive experiments were conducted on a dataset comprising 20 full-match Video clips. The proposed system achieved a root mean square error (RMSE) of 4.87 in possession estimation, outperforming all evaluated baselines, including YOLOv10n (RMSE: 5.12) and YOLOv11 (RMSE: 5.17), with a substantial improvement over YOLOv6n (RMSE: 12.73). These results substantiate the effectiveness of the proposed framework in enhancing the precision, efficiency, and automation of football analytics, offering practical value for coaches, analysts, and sports scientists in professional settings.