Synthetic Occluded Masked Face Recognition using Convolutional Neural Networks

An IEEE conference paper, published in September 2022, on synthetic masked-face data for retraining CNN face recognition. Mask mandates were still breaking systems trained on uncovered faces.

• Real masked-face datasets were scarce in 2022, so we generated the data.
• An OpenCV pipeline maps mask geometry onto existing benchmark face datasets.
• We measured how CNN recognition accuracy responds as coverage and mask design change. Read the paper on IEEE Xplore