Greedy K-means is a variant of the classical K-means, which aims to handle the sensitivity of K-means initialization. Based on the Greedy K-means algorithm (Likas, A., Vlassis, N., & Verbeek, J. J. (2003). The global k-means clustering algorithm. Pattern recognition, 36(2), 451-461), we implement a fast version of Greedy K-means with 59 sampling strategy. Therefore it not only enjoys the theoretical guarantee, but also outperforms other initialization methods in a plenty of data sets.