模型是基于 YOLO26l_seg 的基础上训练，用于眼部的检测与实例分割。

The model is trained based on YOLO26l_seg for eye detection and instance segmentation.