The overall pipeline of the propose

The overall pipeline of the proposed approach is shown in Figure 1, which decomposes the problem of face recog- nition under random partial occlusions into three stages. Stage I: Learning mask generators using the proposed pair- wise differential siamese network (PDSN) to capture the correspondence between occluded facial blocks and cor- rupted feature elements. Stage II: Establishing a mask dic- tionary from the learned mask generators. Stage III: In the testing phase, combining the feature discarding mask (FDM) of random partial occlusions from this dictionary, which is then multiplied with the original feature to elimi- nate the effect of partial occlusions from recognition.