search

StacyYang / gluoncv-torch

PyTorch API for GluonCV Models
MIT License
535 stars 62 forks source link

Pretrained model on cityscapes #12

Open codermckee opened 6 years ago

codermckee commented 6 years ago

Do you have any pretrained model on cityscapes or mapillary?

codermckee commented 6 years ago

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

zhanghang1989 commented 6 years ago

Will add cityscapes in GluonCV in later release.

codermckee commented 6 years ago

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 pretrained model on ImageNet to train object detector and I can get good mAP. But it doesn't work when it comes to semantic segmentation. would you mind giving me some advice? thank you very much!

zhanghang1989 commented 6 years ago

There are many details in the implementation for reproducing SoTA result. Synchronized Batch Normalization is one of the most important key in semantic segmentation.

zhanghang1989 commented 6 years ago

You may start with my repo https://hangzhang.org/PyTorch-Encoding/experiments/segmentation.html

codermckee commented 6 years ago

You may start with my repo https://hangzhang.org/PyTorch-Encoding/experiments/segmentation.html

Thank you very much !

XiaLiPKU commented 6 years ago

@zhanghang1989 Can you also add the pretrained model on COCO? I see that has been published in GluonCV.