The difference between PEG and PEG for detection

[mt-cly]

Hi, thanks for your great work.
I meet some confusion when reading the paper. Specifically, in the part of Supplement C. Example Code, there is a light difference between the two presented Algorithms.
In Algorithm 1 PyTorch snippet of PEG, the PEG includes a Conv layer, while in Algorithm 2PyTorch snippet of PEG for detection, there are additional BN+Relu layers.
I wonder how about the effectiveness comparison of this two setting, would the second setting with BN+Relu be better?
Thank you.