Kim, et al.. FCSS: Fully Convolutional Self-similarity for Dense Semantic Correspondence. IEEE, 2017, doi:10.1109/cvpr.2017.73.