The images in this dataset are all in portrait format, horizontal direction images were not rotated. By following PoseNet (ICCV, 2015) we first rescaled the images horizontally to 256 pixels, then did the random crops of size 224x224.


 author    = {F. Walch and C. Hazirbas and L. Leal-Taixé and T. Sattler and S. Hilsenbeck and D. Cremers},
 title     = {Image-based localization using LSTMs for structured feature correlation},
 booktitle = {ICCV},
 month     = {October},
 year      = {2017},