A Multi-Strategy Integrated Improved YOLOV11 Algorithm and Its Application in Wildlife Object Detection

This study addresses the limitations of the YOLOV11 algorithm in small object detection, loss of original feature information, and prediction box drift by proposing an improved model, YOLOV11-WR. Specifically, to mitigate the issue of information loss in the backbone network, the Convolutional Block Attention Module (CBAM) attention mechanism is incorporated to enhance the model’s focus on small-object features, thereby improving both efficiency and accuracy. To address the insufficient capability in small object detection, the weighted bidirectional feature pyramid network (BiFPN) is introduced to strengthen low-level feature representation and improving multi-scale feature fusion. Furthermore, to resolve prediction box drift caused by the Intersection over Union (IoU) loss function, Unified-IoU (UIoU) loss is adopted as a replacement, thereby enhancing localization accuracy and further optimizing the training effectiveness of the model. To evaluate the effectiveness of the proposed improvements, this study conducted experimental validation on the African-wildlife dataset. The results demonstrate that YOLOV11-WR achieves superior detection performance, with a precision (P) of 0.939, a recall (R) of 0.926, mAP@50 of 0.926, and mAP@50:95 of 0.760. Compared to other models, YOLOV11-WR improves mAP@50 by 0.6%-3% and mAP@50:95 by 1.6%-3%, providing strong evidence of its superior performance in object detection tasks. In practical applications, YOLOV11-WR exhibits significant advantages in detecting small animals such as zebras and antelopes. Compared to existing detection models, YOLOV11-WR is more suitable for real-time monitoring in zoos, wildlife reserves, and nature conservation areas. Furthermore, this model has broader applicability in wildlife conservation research, facilitating population estimation, migration route monitoring, and providing intelligent support for ecological studies and wildlife protection initiatives.