How to understand `padding=35` for the first conv layer?

xwjabc / hed

A PyTorch reimplementation of Holistically-Nested Edge Detection

170 stars 42 forks source link

Closed yang-jin-hai closed 4 years ago

yang-jin-hai commented 4 years ago

I also find this config in the original repo. I think it may be used to increase side outputs resolution, but why it's 35?

yang-jin-hai commented 4 years ago

I figure it out ^_^

debvrat commented 3 years ago

Hi, @WannaSeaU I have the same query. What is the reason behind padding=35 for the first conv layer?

meteorshowers also used padding = 21 for the first conv layer. Why is that? Shouldn't padding be equal to 1 for all the layers?

Also a query for @xwjabc