search

ydhongHIT / DDRNet

The official implementation of "Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes"
MIT License
415 stars 53 forks source link

Apply on 256*256 input #7

Closed zyxu1996 closed 3 years ago

zyxu1996 commented 3 years ago

if the scale of input is 256256, should I rescale it to a large size? Scale 10241024 or something.

ydhongHIT commented 3 years ago

if the scale of input is 256_256, should I rescale it to a large size? Scale 1024_1024 or something.

Since ddrnet is designed for relatively high-resolution image segmentation, you should modify the number of downsampling branches in the dappm module.

geezah commented 2 years ago

Hi, how would you adjust the number of downsampling branches for samples of size 512x512?

songbingyue commented 9 months ago

Hi, how would you adjust the number of downsampling branches for samples of size 512x512?

My image size is also 512x512. Have you solved this problem and how should I set dappm

songbingyue commented 9 months ago

Hi, how would you adjust the number of downsampling branches for samples of size 512x512?

My image size is also 512x512. Have you solved this problem and how should I set dappm