Monitoring waterbirds is vital for evaluating the ecological health of wetlands, and object detection offers an automated solution for identifying birds in monitoring imagery. However, conventional detection methods often overlook the multi-scale nature of bird targets, limiting their ability to capture rich contextual information across different scales. To address this, we propose a cross-dimensional attention network (CDA-Net) for bird detection that integrates spatial and channel information to improve species recognition. The proposed CDA-Net partitions feature maps into multiple channel wise sub-features. Spatial and channel attention are applied to each sub-feature, and the resulting features are fused using the Hadamard product. The fused features are then forwarded to the detection head to generate the final detection results. This approach effectively captures and integrates information across spatial and channel dimensions. Experiments on our self-constructed Nanhai Wetland Waterbird Dataset and the public CUB-200-2011 dataset yield precision scores of 91.32% and 81.99%, respectively, outperforming existing methods. Our approach effectively handles scale variation in bird detection and provides a valuable tool for advancing automated wetland waterbird monitoring.

Wetland waterbirds serve as key ecological indicators for assessing habitat quality and biodiversity. Accurate identification of waterbird species is a cornerstone of long-term ecological monitoring. The resulting data are critical for assessing wetland ecosystem health and biodiversity. However, prevailing recognition approaches often prioritize detection accuracy at the expense of computational efficiency. They are also hindered by complex background heterogeneity and interspecies visual similarity. These limitations hinder the scalability and practical deployment of such methods for on-site ecological monitoring within wetland ecosystems. To address these challenges, this study proposes an optimized end-to-end framework, ShuffleNetV2-iRMB-ShapeIoU-YOLO (SIS-YOLO), designed for robust recognition of wetland waterbirds in complex environments. Specifically, the proposed framework integrates ShuffleNetV2 with inverted Residual Mobile Blocks (iRMB) to improve computational efficiency while maintaining robust feature representation. This design further enables deployment on resource-constrained mobile and embedded platforms. Additionally, ShapeIoU, a refined bounding box similarity metric, is introduced to jointly optimize overlap and shape consistency, effectively mitigating misclassification among visually similar species. Experimental results on the IC-Beijing dataset show that SIS-YOLO achieves 91.1% precision and 79.1% mAP@0.5:0.95 with only 2.9 million parameters. Compared with the lightweight baseline YOLOv8n, it improves precision by 2% and mAP@0.5:0.95 by 1.2%, while requiring fewer parameters and offering higher computational efficiency.

Bird monitoring and protection are essential for maintaining biodiversity, and fine-grained bird classification has become a key focus in this field. Audio-visual modalities provide critical cues for this task, but robust feature extraction and efficient fusion remain major challenges. We introduce a multi-stage fine-grained audiovisual fusion network (MSFG-AVFNet) for fine-grained bird species classification, which addresses these challenges through two key components: (1) the audiovisual feature extraction module, which adopts a multi-stage fine-tuning strategy to provide high-quality unimodal features, laying a solid foundation for modality fusion; (2) the audiovisual feature fusion module, which combines a max pooling aggregation strategy with a novel audiovisual loss function to achieve effective and robust feature fusion. Experiments were conducted on the self-built AVB81 and the publicly available SSW60 datasets, which contain data from 81 and 60 bird species, respectively. Comprehensive experiments demonstrate that our approach achieves notable performance gains, outperforming existing state-of-the-art methods. These results highlight its effectiveness in leveraging audiovisual modalities for fine-grained bird classification and its potential to support ecological monitoring and biodiversity research.

Passive acoustic monitoring (PAM) technology is increasingly becoming one of the mainstream methods for bird monitoring. However, detecting bird audio within complex natural acoustic environments using PAM devices remains a significant challenge. To enhance the accuracy (ACC) of bird audio detection (BAD) and reduce both false negatives and false positives, this study proposes a BAD method based on a Dual-Feature Enhancement Fusion Model (DFEFM). This method incorporates per-channel energy normalization (PCEN) to suppress noise in the input audio and utilizes mel-frequency cepstral coefficients (MFCC) and frequency correlation matrices (FCM) as input features. It achieves deep feature-level fusion of MFCC and FCM on the channel dimension through two independent multi-layer convolutional network branches, and further integrates Spatial and Channel Synergistic Attention (SCSA) and Multi-Head Attention (MHA) modules to enhance the fusion effect of the aforementioned two deep features. Experimental results on the DCASE2018 BAD dataset show that our proposed method achieved an ACC of 91.4% and an AUC value of 0.963, with false negative and false positive rates of 11.36% and 7.40%, respectively, surpassing existing methods. The method also demonstrated detection ACC above 92% and AUC values above 0.987 on datasets from three sites of different natural scenes in Beijing. Testing on the NVIDIA Jetson Nano indicated that the method achieved an ACC of 89.48% when processing an average of 10 s of audio, with a response time of only 0.557 s, showing excellent processing efficiency. This study provides an effective method for filtering non-bird vocalization audio in bird vocalization monitoring devices, which helps to save edge storage and information transmission costs, and has significant application value for wild bird monitoring and ecological research.

Bird vocalizations are pivotal for ecological monitoring, providing insights into biodiversity and ecosystem health. Traditional recognition methods often neglect phase information, resulting in incomplete feature representation. In this paper, we introduce a novel approach to bird vocalization recognition (BVR) that integrates both amplitude and phase information, leading to enhanced species identification. We propose MHAResNet, a deep learning (DL) model that employs residual blocks and a multi-head attention mechanism to capture salient features from logarithmic power (POW), Instantaneous Frequency (IF), and Group Delay (GD) extracted from bird vocalizations. Experiments on three bird vocalization datasets demonstrate our method’s superior performance, achieving accuracy rates of 94%, 98.9%, and 87.1% respectively. These results indicate that our approach provides a more effective representation of bird vocalizations, outperforming existing methods. This integration of phase information in BVR is innovative and significantly advances the field of automatic bird monitoring technology, offering valuable tools for ecological research and conservation efforts.