The MERL system uses 3D face models and specifically targets illumination and pose variation, which they think as the most critical factors limiting performance.

MERL系用三维人脸模型，重点解决他们所认最限制识别率两个因素，光照和变化。