In the process of data annotation, if there are ten vehicles target images, but only five of them for the labeling, five others missing, with this kind of missing target depth data of network training, whether to have influence on the performance of the network? Have influence on which algorithm? For example, yolo, faster - RCNN?