We ablate the key design choices of our method on the CMU Panoptic Studio dataset
[1]. The top video depicts the pre-scan and dynamic sequences on which the models are trained on. Each row shows the result of each ablation on a specific test camera.
Removing each key component significantly drops model performance, highlighting their contribution.
[1] Hanbyul Joo, Hao Liu, Lei Tan, Lin Gui, Bart Nabbe, Iain Matthews, Takeo Kanade, Shohei Nobuhara, Yaser Sheikh. Panoptic Studio: A Massively Multiview System for Social Motion Capture, ICCV 2015