The system, “Occluded Resistant Face Detection and Recognition System”, contains different scale detectors for calculating face locations. Additive angular margin loss is added into the training phase for achieving high efficiency and accuracy. The system can achieves 80% accuracy under the 50% face occluded.

本技術特點在於利用卷積神經網路訓練強大的人臉偵測網路與特徵擷取網路。在人臉遮蔽超過80%時可以正常偵測人臉區域，在臉部遮蔽50%的情況下達到80%辨識準確率，並且達到即時辨識的效果。本技術大幅度減少人臉辨識受到遮蔽的影響。

The technology is applicable to related fields, such as security monitoring and access security. The model training adopts two-stage training, which can quickly train the database, to make the system more convenient and flexible to use.