search

fyu / drn

Dilated Residual Networks
https://www.vis.xyz/pub/drn
BSD 3-Clause "New" or "Revised" License
1.1k stars 219 forks source link

hard to train to see the pole object in the CamVid #15

Closed guanfuchen closed 6 years ago

guanfuchen commented 6 years ago

image

Thanks for your code using drn to segment the CamVid dataset, but I use the code well on most objects segmenting, it is hard to segment the pole object in the CamVid dataset, so what can I do to get the pole object to get segmenting.

fyu commented 6 years ago

What's your training setting? Image resolution?

guanfuchen commented 6 years ago

I use learning rate=1e-5, momentum=0.99, weight_decay=5e-4, and use SGD solver, batch_size=1, Image resolution is 360*480.

-----原始邮件----- 发件人:"Fisher Yu" notifications@github.com 发送时间:2017-12-31 15:18:51 (星期日) 收件人: fyu/drn drn@noreply.github.com 抄送: guanfuchen guanfuchen@zju.edu.cn, Author author@noreply.github.com 主题: Re: [fyu/drn] hard to train to see the pole object in the CamVid (#15)

What's your training setting? Image resolution?

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