To address the limitations of traditional manual egg candling methods, including high labor intensity, low efficiency, and high misjudgment rates, this study proposes an early hatchability detection model for goose eggs based on an improved YOLOv11. The model enhances image feature extraction capabilities by incorporating a P2 small-object detection layer and integrates a deformable local kernel attention (Deformable-LKA) convolution module to achieve dynamic adaptation of detection points to vascular texture features. Tests on a dataset of goose eggs from days 4 to 7 of incubation demonstrate that the improved model achieves a mean average precision (mAP@0.5) of 99.5%, representing a 1.8% improvement over the original model while maintaining a parameter size of 3.45M. Daily testing results show that the model reaches 99.2% mAP@0.5 by day 7 of incubation, effectively balancing performance, speed, and model size. This provides an efficient detection tool for goose egg incubation research.