Code for unsupervised learning of object landmarks as proposed in "Unsupervised Learning of Object Landmarks through Conditional Image Generation", Tomas Jakab*, Ankush Gupta*, Hakan Bilen, Andrea Vedaldi, Advances in Neural Information Processing Systems (NeurIPS) 2018.
Hi, I have a problem with the codes. What is the role of keypoints in the program? If I want to change cubes, do I have to use images with keypoints ?