Scalable Face Image Retrieval with Identity-Based Quantization and Multireference Reranking

By Zhong Wu, Qifa Ke, Jian Sun, and Heung-Yeung Shum

用5个Facial components(两个眼睛、鼻尖、两个嘴角)的检测器代替了Local Point Detector，每个component用5x7的Grid表示。