基于YOLO-DAS模型的湿地水鸟检测方法: 以内蒙古南海子湿地为例

doi:10.17520/biods.2025237

生物多样性 ›› 2025, Vol. 33 ›› Issue (9): 25237. DOI: 10.17520/biods.2025237 cstr: 32101.14.biods.2025237

基于YOLO-DAS模型的湿地水鸟检测方法: 以内蒙古南海子湿地为例

孙济伦¹^,²^,³, 谢将剑¹^,²^,³(), 张长春¹^,²^,³, 张军国¹^,²^,³^,^*()

1.北京林业大学工学院, 北京 100083
2.林木资源高效生产全国重点实验室, 北京 100083
3.北京林业大学生物多样性智慧监测研究中心, 北京 100083

收稿日期:2025-06-20 接受日期:2025-08-08 出版日期:2025-09-20 发布日期:2025-10-31
通讯作者: *E-mail: zhangjunguo@bjfu.edu.cn
基金资助:
国家自然科学基金(32371874);国家自然科学基金(32401569);北京市自然科学基金(5252014);北京林业大学科技创新计划项目(2024XY-G002)

Wetland waterbird detection method based on the YOLO-DAS model: A case study of the Nanhaizi Wetland in Inner Mongolia

Jilun Sun¹^,²^,³, Jiangjian Xie¹^,²^,³(), Changchun Zhang¹^,²^,³, Junguo Zhang¹^,²^,³^,^*()

1 School of Technology, Beijing Forestry University, Beijing 100083, China
2 State Key Laboratory of Efficient Production of Forest Resources, Beijing 100083, China
3 Research Center for Biodiversity Intelligent Monitoring, Beijing Forestry University, Beijing 100083, China

Received:2025-06-20 Accepted:2025-08-08 Online:2025-09-20 Published:2025-10-31
Contact: *E-mail: zhangjunguo@bjfu.edu.cn
Supported by:
National Natural Science Foundation of China(32371874);National Natural Science Foundation of China(32401569);Beijing Natural Science Foundation(5252014);Beijing Forestry University Science and Technology Innovation Program(2024XY-G002)

摘要/Abstract

摘要：

湿地水鸟监测对于生物多样性及湿地保护具有重要意义。随着计算机视觉技术的广泛应用, 利用深度学习模型进行鸟类图像检测已成为鸟类保护的重要手段。实际湿地水鸟监测图像中存在背景信息复杂、类间特征相似、前景遮挡及目标尺度差异等问题, 使得模型检测性能不足。针对以上问题, 本研究建立了包含内蒙古南海子湿地111种水鸟27,030张图像的自建数据集Bird111, 并提出一种基于YOLO-DAS的湿地水鸟目标检测算法。首先, 融合可变形注意力机制(deformable attention, DAT), 自适应地关注图像中的重要区域, 提高网络的特征提取能力, 避免复杂背景以及相似特征的影响; 然后, 利用自适应空间特征融合(adaptively spatial feature fusion, ASFF), 对所提取的不同尺度特征中的冲突信息进行过滤以增强尺度不变性, 提高模型对多尺度鸟类目标的响应能力; 最后, 引入SlideLoss损失函数, 增加训练过程中对困难样本的关注, 提高对小目标和受遮挡目标的检测性能。实验结果表明, YOLO-DAS模型在自建Bird111数据集上相较于其他主流方法拥有最优的检测性能, 其精确率、召回率及平均检测精度均值较基线模型分别提升4%、2.4%和2.9%, 同时在CUB200-2011、Birdsnap和NABirds公开数据集上具有良好的泛化性能。本文所提出的YOLO-DAS模型能够有效提高复杂背景下的小目标或受遮挡鸟类的检测性能, 为湿地水鸟监测工作中不同鸟类目标尺度的图像检测提供了有效的技术方法。

关键词: 湿地水鸟检测, YOLOv8, 可变形注意力, 自适应空间特征融合, 损失函数

Abstract

Aims: Wetland waterbird monitoring is of great significance for biodiversity and wetland conservation. With the widespread application of computer-vision techniques, bird image detection using deep-learning models has become an important tool for bird conservation. However, real wetland monitoring images often contain complex backgrounds, inter-class feature similarity, foreground occlusion, and large target-scale variation, which impair detection performance.

Methods: To overcome these limitations, we built a dataset (Bird111) comprising 27,030 images of 111 waterbird species from the Nanhaizi Wetland, Inner Mongolia, and we propose a wetland waterbird detection algorithm based on YOLO-DAS. First, deformable attention mechanism (DAT) is integrated to adaptively focus on important image regions, improving the network’s feature-extraction ability and reducing the influence of complex backgrounds and similar features. Second, adaptively spatial feature fusion (ASFF) is applied to filter conflicting multi-scale feature information, enhancing scale invariance and the model’s responsiveness to multiscale bird targets. Finally, the SlideLoss function is introduced to increase emphasis on difficult samples during training and to improve detection of small and occluded targets.

Results: Experiments show that the YOLO-DAS model achieves the best detection performance on the Bird111 dataset compared with other mainstream methods. Its precision, recall, and mean average precision improved by 4.0%, 2.4%, and 2.9%, respectively, relative to the baseline model. The model also generalized well to public datasets (CUB- 200-2011, Birdsnap, and NABirds).

Conclusion: The YOLO-DAS model proposed here can effectively improve detection of small or occluded birds in complex backgrounds, and provides a practical technical approach for multi-scale bird detection in wetland monitoring.

Key words: wetland waterbird detection, YOLOv8, deformable attention, adaptively spatial feature fusion, loss function

孙济伦, 谢将剑, 张长春, 张军国 (2025) 基于YOLO-DAS模型的湿地水鸟检测方法: 以内蒙古南海子湿地为例. 生物多样性, 33, 25237. DOI: 10.17520/biods.2025237.

Jilun Sun, Jiangjian Xie, Changchun Zhang, Junguo Zhang (2025) Wetland waterbird detection method based on the YOLO-DAS model: A case study of the Nanhaizi Wetland in Inner Mongolia. Biodiversity Science, 33, 25237. DOI: 10.17520/biods.2025237.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.biodiversity-science.net/CN/10.17520/biods.2025237

https://www.biodiversity-science.net/CN/Y2025/V33/I9/25237

图/表 16

参考文献 45

[1]	Adarsh P, Rathi P, Kumar M (2020) YOLO v3-Tiny: Object detection and recognition using one stage improved model. In: 2020 International Conference on Advanced Computing and Communication Systems (ICACCS), pp. 687-694. IEEE, Coimbatore.
[2]	Berg T, Liu JX, Lee SW, Alexander ML, Jacobs DW, Belhumeur PN (2014) Birdsnap: Large-scale fine-grained visual categorization of birds. In: IEEE 2014 Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2019-2026. IEEE, Columbus.
[3]	Chalmers C, Fergus P, Wich S, Longmore SN, Walsh ND, Stephens PA, Sutherland C, Matthews N, Mudde J, Nuseibeh A (2023) Removing human bottlenecks in bird classification using camera trap images and deep learning. Remote Sensing, 15, 2638. DOI URL
[4]	Dong XY, Li ZT, Liu YP, Gao L, Li RX, Liu L (2023) Structure and diversity of bird community of Nanhaizi Wetland Nature Reserve in Baotou. Journal of Inner Mongolia Normal University (Natural Science Edition), 52, 413-420. (in Chinese with English abstract)
	[董向阳, 李子腾, 刘云鹏, 高丽, 李润煊, 刘利 (2023) 包头南海子湿地自然保护区鸟类群落结构及多样性研究. 内蒙古师范大学学报(自然科学汉文版), 52, 413-420.
[5]	Fan JC, Liu XX, Wang XZ, Wang DY, Han M (2020) Multi-background island bird detection based on faster R-CNN. Cybernetics and Systems, 52, 26-35. DOI URL
[6]	Gao DZ, Lin H, Lin LL, Cui GF (2021) The feasibility of wetland waterfowl monitoring method based on UAV remote sensing. Chinese Journal of Zoology, 56, 100-110. (in Chinese with English abstract)
	[高大中, 林海, 林乐乐, 崔国发 (2021) 利用小型无人机监测西洞庭湖水鸟的可行性探讨. 动物学杂志, 56, 100-110.]
[7]	Jocher G (2023a) YOLOv5 Release v6.0. https://github.com/ultralytics/yolov5/releases/tag/v6.0. (accessed on 2023-01-22)
[8]	Jocher G (2023b) YOLOv8 Release v8.1.0. https://github.com/ultralytics/ultralytics/releases/tag/v8.1.0. (accessed on 2023- 11-10)
[9]	Jocher G, Qiu J (2024) YOLOv11 Release v11.0.0. https://github.com/ultralytics/ultralytics. (accessed on 2024-12-17)
[10]	Kassim YM, Byrne ME, Burch C, Mote K, Hardin J, Larsen DR, Palaniappan K (2020) Small object bird detection in infrared drone videos using mask R-CNN deep learning. Electronic Imaging, 32, 1-8.
[11]	Li BC, Zhang JG, Zhang CC, Wang LF, Xu JL, Liu L (2024) Rare bird recognition method in Beijing based on TC-YOLO model. Biodiversity Science, 32, 24056. (in Chinese with English abstract) DOI
	[李柏灿, 张军国, 张长春, 王丽凤, 徐基良, 刘利 (2024) 基于TC-YOLO模型的北京珍稀鸟类识别方法. 生物多样性, 32, 24056.] DOI
[12]	Li C, Zhang BC, Hu HW, Dai J (2019) Enhanced bird detection from low-resolution aerial image using deep neural networks. Neural Processing Letters, 49, 1021-1039. DOI
[13]	Li CM, Chen JS, Liao XL, Ramus AP, Angelini C, Liu LL, Silliman BR, Bertness MD, He Q (2023) Shorebirds-driven trophic cascade helps restore coastal wetland multifunctionality. Nature Communications, 14, 8076. DOI PMID
[14]	Liu JX, Kang BY, Liu C, Peng XH, Bai Y (2024) YOLO-BFRV: An efficient model for detecting printed circuit board defects. Sensors, 24, 6055. DOI URL
[15]	Liu SL, Li YL, Qu JY, Wu RB (2022) Airport UAV and birds detection based on deformable DETR. Journal of Physics: Conference Series, 2253, 012024. DOI
[16]	Liu ST, Huang D, Wang YH (2019) Learning spatial fusion for single-shot object detection. arXiv, doi: 10.48550/arXiv.1911.09516.
[17]	Liu W, Anguelov D, Erhan D, Szegedy C, Reed S, Fu CY, Berg AC (2016) SSD:Single shot MultiBox detector. In: 2016 European Conference on Computer Vision (ECCV), pp. 21-37. Springer International Publishing, Amsterdam.
[18]	Luo YL, Pan YH, Xia SX, Tao YZ (2024) Lightweight apple leaf disease detection algorithm based on improved YOLOv8. Smart Agriculture, 6(5), 128-138. (in Chinese with English abstract)
	[罗友璐, 潘勇浩, 夏顺兴, 陶友志 (2024) 基于改进YOLOv8的苹果叶病害轻量化检测算法. 智慧农业(中英文), 6(5), 128-138.]
[19]	Ma CW, Zhang H, Ma XM, Wang JL, Zhang YS, Zhang XA (2024) Method for the lightweight detection of wheat disease using improved YOLOv8. Transactions of the Chinese Society of Agricultural Engineering, 40(5), 187-195. (in Chinese with English abstract)
	[马超伟, 张浩, 马新明, 王键霖, 张永爽, 张小艾 (2024) 基于改进YOLOv8的轻量化小麦病害检测方法. 农业工程学报, 40(5), 187-195.]
[20]	Ma XL, Wang ML, Kuang HL, Tang L, Liu XH (2024) Detecting and counting silkworms using improved YOLOv8n. Transactions of the Chinese Society of Agricultural Engineering, 40(15), 143-151. (in Chinese with English abstract)
	[马小林, 王梦麟, 旷海兰, 唐亮, 刘新华 (2024) 基于YOLOv8n改进的蚕虫检测与计数方法. 农业工程学报, 40(15), 143-151.]
[21]	Mpouziotas D, Karvelis P, Tsoulos I, Stylios C (2023) Automated wildlife bird detection from drone footage using computer vision techniques. Applied Sciences, 13, 7787. DOI URL
[22]	Muhammad SI, Junior HH, Ringim AS, Muhammad IL, Onoja J (2022) Waterbird population estimates in Hadejia-Nguru wetlands: Analysis of a five-year monitoring program. Wetlands, 42, 12. DOI
[23]	Ren SQ, He KM, Girshick R, Sun J (2016) Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39, 1137-1149. DOI URL
[24]	Song Q, Guan Y, Guo X, Guo XH, Chen YF, Wang HF, Ge JP, Wang TM, Bao L (2024) Benchmarking wild bird detection in complex forest scenes. Ecological Informatics, 80, 102466. DOI URL
[25]	Song ZY, Yang KH, Zhang Y (2022) Bird detection algorithm in natural scenes based on improved YOLOv3. Laser & Optoelectronics Progress, 59(18), 1815013. (in Chinese with English abstract)
	[宋子盈, 杨奎河, 张宇 (2022) 基于改进YOLOv3的自然场景下鸟类检测算法. 激光与光电子学进展, 59(18), 1815013.]
[26]	Talaat FM, ZainEldin H (2023) An improved fire detection approach based on YOLO-v8 for smart cities. Neural Computing and Applications, 35, 20939-20954. DOI
[27]	Te XT, Xie B (2022) Composition and diversity of birds excluding Passeriformes in Inner Mongolia Hulun Lake National Nature Reserve in 2017. Wetland Science, 20, 483-489. (in Chinese with English abstract)
	[特喜铁, 谢宾 (2022) 2017年内蒙古呼伦湖国家级自然保护区非雀形目鸟类组成和多样性. 湿地科学, 20, 483-489.]
[28]	Tie J, Zhu CG, Zheng L, Wang HJ, Ruan CW, Wu M, Xu K, Liu JQ (2024) LSKA-YOLOv8: A lightweight steel surface defect detection algorithm based on YOLOv8 improvement. Alexandria Engineering Journal, 109, 201-212. DOI URL
[29]	Tong FC, Zheng ZW, Lin JL, Huang ZJ, Yang XZ, Wu MF, Zhang XL, Xiao YH (2023) Bird diversity of Maofeng Mountain Forest Park in Guangzhou. Tropical Geography, 43, 1726-1737. (in Chinese with English abstract) DOI
	[佟富春, 郑泽惟, 林嘉莉, 黄子峻, 杨玄宗, 吴牧凡, 张晓玲, 肖以华 (2023) 广州市帽峰山森林公园鸟类多样性研究. 热带地理, 43, 1726-1737.] DOI
[30]	Van Horn G, Branson S, Farrell R, Haber S, Barry J, Ipeirotis P, Perona P, Belongie S (2015) Building a bird recognition app and large scale dataset with citizen scientists:The fine print in fine-grained dataset collection. In: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 595-604. IEEE, Boston.
[31]	Varghese R, Sambath M (2024) YOLOv8: A novel object detection algorithm with enhanced performance and robustness. In: 2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems (ADICS), pp. 1-6. IEEE, Chennai.
[32]	Vo HT, Thien NN, Mui KC (2023) Bird detection and species classification: Using YOLOv5 and deep transfer learning models. International Journal of Advanced Computer Science and Applications, 14, 939-947.
[33]	Wah C, Branson S, Welinder P, Perona P, Belongie S (2011) The Caltech-UCSD Birds-200-2011 Dataset, Technical Report CNS-TR-2011-001. California Institute of Technology, California, USA.
[34]	Wang CX, Zhang ZW, Xia SX, Duan HL, Wang W, Jia YF, Zhang LX, Feng G, Yang YQ, Li T, Ding CQ, Wang CP, Yuan BD, Lei JY, Liu Y, Shi JB, Lan KQ, Shi QQ, Xiao Q, Yu XB (2024) Seasonal and regional patterns and conservation strategies of waterbird diversity in the Yellow River Basin. Biodiversity Science, 32, 23490. (in Chinese with English abstract) DOI
	[王春晓, 张正旺, 夏少霞, 段后浪, 王稳, 贾亦飞, 张立勋, 冯刚, 杨亚桥, 李桐, 丁长青, 王春平, 原宝东, 雷进宇, 刘宇, 石建斌, 兰科其, 石青青, 肖晴, 于秀波 (2024) 黄河流域水鸟多样性季节和区域特征及保护策略. 生物多样性, 32, 23490.] DOI
[35]	Wang HC, Xia F, Zhang YY, Liu YJ, Liu S, Song F, Jian HF (2024) Bird and habitat recognition based on deep learning algorithm: A case study of Beijing Cuihu National Urban Wetland Park. Chinese Journal of Ecology, 43, 2231-2238. (in Chinese with English abstract) DOI
	[王洪昌, 夏舫, 张渊媛, 刘颖杰, 刘松, 宋飞, 鉴海防 (2024) 基于深度学习算法的鸟类及其栖息地识别——以北京翠湖国家城市湿地公园为例. 生态学杂志, 43, 2231-2238.]
[36]	Wang XQ, Gao HB, Jia ZM, Li ZJ (2023) BL-YOLOv8: An improved road defect detection model based on YOLOv8. Sensors, 23, 8361.
[37]	Wang Y, Zhou JG, Zhang CY, Luo ZP, Han XX, Ji YZ, Guan JH (2023) Bird object detection: Dataset construction, model performance evaluation, and model lightweighting. Animals, 13, 2924. DOI URL
[38]	Wu ET, Wang HC, Lu HX, Zhu WQ, Jia YF, Wen L, Choi CY, Guo HM, Li B, Sun LL, Lei GC, Lei JL, Jian HF (2022) Unlocking the potential of deep learning for migratory waterbirds monitoring using surveillance video. Remote Sensing, 14, 514. DOI URL
[39]	Wu XY, Guo W, Zhu YP, Zhu HJ, Wu HR (2024) Transplant status detection algorithm of cabbage in the field based on improved YOLOv8s. Smart Agriculture (in Chinese and English), 6(2), 107-117. (in Chinese with English abstract)
	[吴小燕, 郭威, 朱轶萍, 朱华吉, 吴华瑞 (2024) 基于改进YOLOv8s的大田甘蓝移栽状态检测算法. 智慧农业(中英文), 6(2), 107-117.]
[40]	Xia ZF, Pan XR, Song SJ, Li LE, Huang G (2022) Vision transformer with deformable attention. In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 4794-4803. IEEE, New Orleans.
[41]	Xie J, Deng YM, Wang RM (2024) Improved traffic sign detection algorithm based on YOLOv8s. Computer Engineering, 50(11), 338-349. (in Chinese with English abstract) DOI
	[谢竞, 邓月明, 王润民 (2024) 改进YOLOv8s的交通标志检测算法. 计算机工程, 50(11), 338-349.] DOI
[42]	Yang XX, Li SB, Chen TE, Xie ZH (2023) Bird diversity of Baigui Lake Natural Reserve in winter. Journal of Ecology and Rural Environment, 39, 90-96. (in Chinese with English abstract)
	[杨晓星, 李少斌, 陈天恩, 谢朝晖 (2023) 白龟湖自然保护区冬季鸟类多样性分析. 生态与农村环境学报, 39, 90-96.]
[43]	Yu ZP, Huang HB, Chen WJ, Su YX, Liu YH, Wang XY (2024) YOLO-FaceV2: A scale and occlusion aware face detector. Pattern Recognition, 155, 110714. DOI URL
[44]	Zhang LF, Tian Y (2024) Improved YOLOv8 multi-scale and lightweight vehicle object detection algorithm. Computer Engineering and Applications, 60(3), 129-137. (in Chinese with English abstract) DOI
	[张利丰, 田莹 (2024) 改进YOLOv8的多尺度轻量型车辆目标检测算法. 计算机工程与应用, 60(3), 129-137.] DOI
[45]	Zhu HY, Cheng HX, Luo XL (2024) Application of improved YOLOv8 network in insulator defect detection. Water Resources and Power, 42(5), 183-187. (in Chinese with English abstract)
	[朱泓宇, 程换新, 骆晓玲 (2024) 改进YOLOv8网络在绝缘子缺陷检测中的应用. 水电能源科学, 42(5), 183-187.]

超参数 Hyper-parameters	设定值 Setting value
学习率 Learning rate	0.005
图像尺寸 Image size	640 × 640
优化器 Optimizer	SGD
批次大小 Batch size	32
线程数 Workers	16
最大迭代次数 Maximum epochs	300
权重衰减率 Weight decay	0.0005

超参数 Hyper-parameters	设定值 Setting value
学习率 Learning rate	0.005
图像尺寸 Image size	640 × 640
优化器 Optimizer	SGD
批次大小 Batch size	32
线程数 Workers	16
最大迭代次数 Maximum epochs	300
权重衰减率 Weight decay	0.0005

可变形注意力机制 Deformable attention transformer (DAT)	自适应特征融合机制 Adaptively structure feature fusion (ASFF)	滑动损失函数 SlideLoss	精确率 Precision (P, %)	召回率 Recall (R, %)	平均精度均值 Mean average precision (mAP50, %)	浮点计算量 Giga floating-point operations per second (GFLOPS)
-	-	-	73.4	75.5	78.3	9.9
√	-	-	77.0	77.9	80.1	10.0
-	√	-	78.8	77.1	81.0	11.9
-	-	√	77.8	76.4	80.3	9.9
√	√	-	79.7	77.3	81.2	12.2
√	-	√	78.8	76.6	80.5	10.0
-	√	√	77.7	76.3	80.6	11.9
√	√	√	77.4	77.9	81.2	12.2

可变形注意力机制 Deformable attention transformer (DAT)	自适应特征融合机制 Adaptively structure feature fusion (ASFF)	滑动损失函数 SlideLoss	精确率 Precision (P, %)	召回率 Recall (R, %)	平均精度均值 Mean average precision (mAP50, %)	浮点计算量 Giga floating-point operations per second (GFLOPS)
-	-	-	73.4	75.5	78.3	9.9
√	-	-	77.0	77.9	80.1	10.0
-	√	-	78.8	77.1	81.0	11.9
-	-	√	77.8	76.4	80.3	9.9
√	√	-	79.7	77.3	81.2	12.2
√	-	√	78.8	76.6	80.5	10.0
-	√	√	77.7	76.3	80.6	11.9
√	√	√	77.4	77.9	81.2	12.2

模型 Model	精确率 Precision (P, %)	召回率 Recall (R, %)	F1分数 F1-score (F1, %)	平均精度均值 Mean average precision (mAP50, %)	浮点计算量 Giga floating-point operations per second (GFLOPS)	参考文献 References
Faster R-CNN	71.2	70.5	70.8	72.6	124.8	Ren et al, 2016
SSD	75.2	75.3	75.3	77.8	15.6	Liu et al, 2016
YOLOv3-Tiny	74.6	69.3	71.9	75.0	19.0	Adarsh et al, 2020
YOLOv5s	78.6	76.1	77.3	79.1	16.9	Jocher, 2023a
YOLOv11n	76.8	75.5	75.1	77.5	6.8	Jocher & Qiu, 2024
YOLOv8n	73.4	75.5	74.4	78.3	9.9	Jocher, 2023b
YOLO-DAS	77.4	77.9	77.7	81.2	12.2	本文 This study

基于YOLO-DAS模型的湿地水鸟检测方法: 以内蒙古南海子湿地为例

Wetland waterbird detection method based on the YOLO-DAS model: A case study of the Nanhaizi Wetland in Inner Mongolia

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 16

参考文献 45

相关文章 1

编辑推荐

Metrics

本文评价

模型 Model	精确率 Precision (P, %)	召回率 Recall (R, %)	F1分数 F1-score (F1, %)	平均精度均值 Mean average precision (mAP50, %)	浮点计算量 Giga floating-point operations per second (GFLOPS)	参考文献 References
YOLOv8n + Triplet	76.6	76.9	76.8	79.0	9.9	张利丰和田莹, 2024
YOLOv8n + LSKA	76.5	75.7	76.1	79.3	10.0	Tie et al, 2024
YOLOv8n + MSDA	76.8	76.9	76.8	79.5	10.4	罗友璐等, 2024
YOLOv8n + ACmix	77.1	76.2	76.6	79.5	10.4	朱泓宇等, 2024
YOLOv8n + RCS-OSA	77.0	77.2	77.1	79.7	21.3	马小林等, 2024
YOLOv8n + DAT	77.0	77.9	77.5	80.1	10.0	本文 This study

模型 Model	精确率 Precision (P, %)	召回率 Recall (R, %)	F1分数 F1-score (F1, %)	平均精度均值 Mean average precision (mAP50, %)	参考文献 References
YOLOv8n + Focal Loss	72.2	68.1	70.1	73.9	谢竞等, 2024
YOLOv8n + Varifocal Loss	74.8	76.3	75.6	77.5	Liu et al, 2024
YOLOv8n + Quality Focal Loss	73.0	73.6	73.3	77.6	吴小燕等, 2024
YOLOv8n + CIoU	76.3	75.9	76.1	79.0	Jocher, 2023b
YOLOv8n + SIoU	76.8	76.0	76.4	79.2	Jocher, 2023b
YOLOv8n + SlideLoss	77.8	76.4	77.1	80.3	本文 This study

数据集 Dataset	模型 Model	精确率 Precision (P, %)	召回率 Recall (R, %)	平均精度均值 Mean average precision (mAP50, %)
CUB200-2011	YOLOv8n	77.5	74.5	83.4
CUB200-2011	YOLO-DAS	79.1	76.7	84.3
Birdsnap	YOLOv8n	65.4	64.8	70.7
Birdsnap	YOLO-DAS	68.0	66.3	73.0
NABirds	YOLOv8n	76.1	73.3	80.4
NABirds	YOLO-DAS	80.2	77.2	84.3

保护等级 Conservation status	物种名称 Species name	平均精度均值 Mean average precision (mAP50, %)
保护等级 Conservation status	物种名称 Species name	YOLOv8n		YOLO-DAS
国家一级 National Grade I	青头潜鸭 Aythya baeri	94.8		95.2 (↑0.4 )
	黑鹳 Ciconia nigra	53.7		72.6 (↑18.9 )
	丹顶鹤 Grus japonensis	69.1		69.6 (↑0.5 )
	白鹤 Grus leucogeranus	72.8		73.4 (↑0.6 )
	遗鸥 Larus relictus	94.5		96.1 (↑1.6 )
	中华秋沙鸭 Mergus squamatus	88.3		88.4 (↑0.1 )
	卷羽鹈鹕 Pelecanus crispus	94.9		94.6 (↓0.3 )
国家二级 National Grade II	鸳鸯 Aix galericulata	92.7		96.4 (↑3.7 )
	鸿雁 Anser cygnoides	66.7		67.0 (↑0.3 )
	蓑羽鹤 Anthropoides virgo	92.1		94.4 (↑2.3 )
	翻石鹬 Arenaria interpres	96.3		96.7 (↑0.4 )
	小天鹅 Cygnus columbianus	55.4		62.8 (↑7.4 )
	大天鹅 Cygnus cygnus	63.7		66.8 (↑3.1 )
	疣鼻天鹅 Cygnus olor	72.8		73.2 (↑0.4 )
	灰鹤 Grus grus	85.1		88.1 (↑3.0 )
	阔嘴鹬 Limicola falcinellus	92.0		93.7 (↑1.7 )
	斑头秋沙鸭 Mergellus albellus	90.2		91.9 (↑1.7 )
	白腰杓鹬 Numenius arquata	58.7		61.0 (↑2.3 )
	大杓鹬 Numenius madagascariensis	71.1		72.7 (↑1.6 )
	小杓鹬 Numenius minutus	65.2		66.3 (↑1.1 )
	鹗 Pandion haliaetus	97.5		98.0 (↑0.5 )
	白琵鹭 Platalea leucorodia	91.5		91.8 (↑0.3 )
	角䴙䴘 Podiceps auritus	90.9		91.2 (↑0.3 )
	黑颈䴙䴘 Podiceps nigricollis	92.8	92.3 (↓0.5 )
	花脸鸭 Sibirionetta formosa	91.7	92.1 (↑0.4 )
	小青脚鹬 Tringa guttifer	80.5	82.0 (↑1.5 )