Fail to reproduce the reported results

[caddyless]

Hi, I run the code but fail to reproduce the reported results on NYU_depth. Specifically, I follow the instruction to prepare data and manage to reproduce the baseline results. However, when I run hourglass with the given configuration ($\alpha$=6, num_cluster=576), the output results are far poor than the reported ones (2.86 vs 0.360). I hope some insights from you.

[PkuRainBow]

@kxqt