Calculation complexity and evaluation on HILTI dataset

[engcang]

Hi.
Thank you for sharing your great work.

1. I've tried SLICT1.0 on NTU-VIRAL dataset and found the results are great except tnp_02 which shows divergence all the trials. Do you have any guess on this?
2. With i9-14900K CPU, SLICT1.0 consumes about ~975% of CPU among 3200% and shows ~150ms. With i9-10900K CPU, it takes ~800ms on HILTI2021 dataset, which cannot be considered as real-time operation. Is this normal you think?
3. I've also tried SLICT1.0 on HILTI2021 dataset (Basement 1, Basement 4, Campus 2, Construction Site 2, LAB Survey 2, UZH Tracking Area Run 2) and the half of the results show divergence (specifically, on Basement 1, Campus 2, Construction Site 2). Have you ever tried SLICT1.0 on HILTI2021 dataset?