Thanks to this, I was able to get a grasp of the 3D HUMANPOSE project. It seems like anyone with some understanding of computer vision can follow along, and the paper review really helps by explaining the performance improvement directions and the logic behind it before diving into implementation, rather than just blindly coding it.