Why the pool_size is set to (20, 12) in the SPHead?

houqb / SPNet

Code for our CVPR2020 paper "Strip Pooling: Rethinking Spatial Pooling for Scene Parsing"

MIT License

387 stars 57 forks source link

Why the pool_size is set to (20, 12) in the SPHead? #9

Open yiminglin-ai opened 4 years ago

yiminglin-ai commented 4 years ago

Hi, Thank you for open-sourcing the amazing work. In the SPHead module, the pool_size parameter er is set to 20, 12.. Can I ask why these two values? If the input image is 480x480 as in ade20k, the feature map produced by the ResNet backbone is 15x15 (32 down) which is smaller than 20x20. It actually becomes an upsampling operation rather than pooling.

VaingloryD commented 4 years ago

same question，+1

yangdonghan50 commented 4 years ago

@ylin-ai 'we improve the original ResNet with the dilation strategy and the final feature map size is set to 1/8 of the input image'

David-Chenn commented 3 years ago

@yangdonghan50 hi, do you know how the pooling sizes 20 and 12 are inferred? or the number 20 and 12 were just set to the given proportion of 480? Thanks !!!