Unsupervised Disentanglement of Pose, Appearance and Background from Images and Videos

Aysegul Dundar, Kevin J. Shih, Animesh Garg, Robert Pottorf, Andrew Tao, Bryan Catanzaro

This checkpoint provides a pretrained model on the BBCPose dataset for learning unsupervised landmarks from video data using a foreground-background factorized image reconstruction. Details are provided in the following tech report: https://arxiv.org/abs/2001.09518

The source code to accompany this model is located at https://github.com/NVIDIA/UnsupervisedLandmarkLearning