search

Dootmaan / DSRL

Implementation of CVPR 2020 Dual Super-Resolution Learning for Semantic Segmentation
MIT License
95 stars 17 forks source link

about input size #14

Open Ontheroad123 opened 3 years ago

Ontheroad123 commented 3 years ago

I see the model's input reduce by half, SSSR and SISR output's size is origin input size, i think the output size is twice bigger than origin input size.

Dootmaan commented 3 years ago

Hi there, sry for the long time waiting. I have been working on another project so i didn't pay much attention on this repo recent days.

Yes thats basically what DSRL is doing, generating a 2x larger output image than the input. I'm not sure I understand what you are trying to say but i guess that you mean the input should use the original size image and the output should be twice larger than it. Well you can always do this as long as you have the HR ground truth masks. Actually it is no big deal for the experiment itself because the size of input and output is always relative.

Ontheroad123 commented 3 years ago

thankyou , i have understand what you said

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2021年7月12日(星期一) 晚上7:56 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [Dootmaan/DSRL] about input size (#14)

Hi there, sry for the long time waiting. I have been working on another project so i didn't pay much attention on this repo recent days.

Yes thats basically what DSRL is doing, generating a 2x larger output image than the input. I'm not sure I understand what you are trying to say but i guess that you mean the input should use the original size image and the output should be twice larger than it. Well you can always do this as long as you have the HR ground truth masks: the size of input and output is always relative.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.