do CVPR paper's model consist of SE block?

[gengshi]

we see that there is a SE block on CVPR2019 at Figure 2, but not seen at the figure of this repo as well as 'resnetDLAS_A' on resnet_caffe_DLAS.py. I wonder if the performance shown at paper Table 3 of line:
"A. resnet-34 with MLFA | CASIA-Web face [25] | attack 3-fold | 99.87" including SE block, or, in other words, which architecture from the code is 'resnet-34 with MLFA ', is that 'resnetDLAS_A' ?

[gengshi]

because I don't see any model in the repo is the same as described on the CVPR paper, hope for reply, thanks!

[joemathai]

Hi, Can you please provide the exact models used for the CVPR paper results? It would really help if you could provide the path to your pre-trained models too.