Almost all existing amodal segmentation methods make the inferences of occluded regions by using features corresponding to the whole image.

翻译：几乎所有现有的现代分解方法都使用与整个图像相应的特征来推断隐蔽区域。