This thesis presents an AI system that uses YOLOv8 to automatically identify incubation and shading behaviors in the Madagascar plover. The model was trained on expert-annotated nest videos and achieved high accuracy when compared with manual observations. It worked especially well in videos with stable lighting and clear bird postures. Using the AI system also saved time, reducing annotation work by almost 30%. These results show that deep learning can help researchers study bird behavior more efficiently. With more data and additional behaviors, the system could become an even more useful tool for conservation research.