codermckee

@Cynicsss Hi, how long did it take you to reach 70% mIoU? I tried the code, but it takes one night just training about 12000 iterations.

@Cynicsss I use 3 *1080ti to train on CityScapes without any change, but just reaching 70.23% mIoU. Is there something I should change in order to reach higher mIoU? BTW,...

@seeyouagain111 老铁， 你训练12个小时能迭代多少次？我只能迭代10000多次， 作者怎么13个小时能训练41000次？我也是2块1080ti。

@seeyouagain111 iter: 21330/41000, lr: 0.005279, loss: 1.0521, eta: 1 day, 4:20:24, time: 24.8223 iter: 21335/41000, lr: 0.005278, loss: 0.8442, eta: 1 day, 4:19:57, time: 24.8530 iter: 21335/41000, lr: 0.005278, loss:...

> Hi, i have the same problem. Which pretrained model did you use that you sloved this issue? @FeiWard I also met this problem. the loss never decrease.

update. Now I use the cityscapes-model as pretrained model to train mapillary project of 66 classes. the loss dropped to 0.1～2.

I use the models pretrained on ImageNet or voc to train mapillary dataset, but get low mIoU. so sad.

> Will add cityscapes in GluonCV in later release. thanks for your reply. I am a beginner of semantic segmentation, and I have a problem which confused me. I use...

> You may start with my repo https://hangzhang.org/PyTorch-Encoding/experiments/segmentation.html Thank you very much !

我也有同样的问题，比自带的Dataset慢，请问你后来解决了吗？