부산대학교 도서관

Simple Summary: We advanced a deep learning model that significantly enhances the detection and population estimation of feral pigeons in the dynamic urban landscape of Hong Kong, employing computer vision techniques. The inherent challenges associated with pigeon concealment within complex urban structures and their high mobility necessitate a robust and effective strategy. Our improved model, Swin-Mask R-CNN with SAHI, integrates a Swin Transformer network for deep feature extraction, a feature pyramid network to enhance multi-scale learning, and three distinct detection heads for classification, bounding box prediction, and segmentation of feral pigeons, respectively. With the assistance of the Slicing-Aided Hyper Inference tool (SAHI), our model excels at detecting small-target pigeons in high-resolution images. Experimental results have demonstrated a substantial 10% increase in AP50s (average precision at 50% intersection over union) compared to the Mask R-CNN approach. This improvement signifies the immense potential of our model in dynamic pigeon detection and accurate population estimation. The success of our novel approach provides a promising solution for effectively managing urban wildlife populations. The overpopulation of feral pigeons in Hong Kong has significantly disrupted the urban ecosystem, highlighting the urgent need for effective strategies to control their population. In general, control measures should be implemented and re-evaluated periodically following accurate estimations of the feral pigeon population in the concerned regions, which, however, is very difficult in urban environments due to the concealment and mobility of pigeons within complex building structures. With the advances in deep learning, computer vision can be a promising tool for pigeon monitoring and population estimation but has not been well investigated so far. Therefore, we propose an improved deep learning model (Swin-Mask R-CNN with SAHI) for feral pigeon detection. Our model consists of three parts. Firstly, the Swin Transformer network (STN) extracts deep feature information. Secondly, the Feature Pyramid Network (FPN) fuses multi-scale features to learn at different scales. Lastly, the model's three head branches are responsible for classification, best bounding box prediction, and segmentation. During the prediction phase, we utilize a Slicing-Aided Hyper Inference (SAHI) tool to focus on the feature information of small feral pigeon targets. Experiments were conducted on a feral pigeon dataset to evaluate model performance. The results reveal that our model achieves excellent recognition performance for feral pigeons. [ABSTRACT FROM AUTHOR]