半監(jiān)督式野生動物夜間目標端到端檢測
打開文本圖片集
End-to-end recognition nighttime wildlife based on semi-supervised learning
LU Han1，CUI Bolun2，WAN Huayang1， ZHANG Gueng1，SHEN Chen1，WANG Chil*
（1. School ， ， ，
2. & Electricity， lOoo94，China） *Correspondingauthor，E-mail：wangchi@shu.edu.cn
Abstract： This study addresses the challenges low accuracy efficiency in the detection wildlife at night，as well as the dificulties associated with manual comprehensive labeling.An end-to-end recognition model for nightime wildlife based on semi-supervised learning（SAN-YOLO）was proposed investigated.A feature atention mechanism a pixel attention mechanism were integrated within the YOLOv8 framework to enhance the adaptability feature representation capabilities the detector for nocturnal mages. Subsequently，a semi-supervised training network based on a teacher-student learning paradigm was constructed，allowing the student model to learn from a substantial number unlabeled original images by generating appropriately asigning pseudo-labels. The efficacy the constructed dataset was then evaluated. Experimental results demonstrate that the mean Average Precision （mAP） SAN
YOLO reaches 69.7% with only 5% annotated data，surpassing the 59.6% mAP achieved with full su pervision in its conventional detector exceeding the baseline model's performance 57.1% . Consequently，the proposed detection method exhibits robust performance with a limited number labeled datasets for nocturnal animals validates the effectiveness attention mechanisms in the domain nighttime object detection.
Key words： object detection；semi-supervised learning；infrared nightvision；wildlife conservation; teacher-student model； attention mechanism
1引言
野生動物是寶貴的自然資源，在生態(tài)系統(tǒng)中發(fā)揮著重要作用1，開展長期、精準且系統(tǒng)的野生動物監(jiān)測對于實施科學的保護策略具有重要意義。（剩余14921字）