国产欧美精品一区二区,中文字幕专区在线亚洲,国产精品美女网站在线观看,艾秋果冻传媒2021精品,在线免费一区二区,久久久久久青草大香综合精品,日韩美aaa特级毛片,欧美成人精品午夜免费影视

Aiming at the low efficiency of traditional face alignment algorithm, a face index based Gaussian difference (DoG) feature and Gaussian process regression tree (GPRT) face key detection algorithm are proposed. First, the kernel between the two inputs is measured by the kernel of the Gaussian process regression tree and is expressed as the number of trees entering the same leaf for both inputs. Then, the shape index DoG feature is extracted based on the Gaussian process regression tree model, and the feature design of the GPRT is further completed. Finally, the filtering response is collected from the local retina mode to increase the stability and achieve robustness against geometric differences. The verification results on the LFPW face database show that the proposed method can achieve good performance, and proves the validity of the shape index-based DoG feature and GPRT face key detection algorithm.