search

irfanICMLL / structure_knowledge_distillation

The official code for the paper 'Structured Knowledge Distillation for Semantic Segmentation'. (CVPR 2019 ORAL) and extension to other tasks.
BSD 2-Clause "Simplified" License
694 stars 104 forks source link

I can't get your mIoU on cityscapes dataset. #29

Closed mailhoho closed 4 years ago

mailhoho commented 4 years ago

Hi, I trained your script PSPNet-resnet18. (student baseline) But, I can't get refered on paper mIoU(69.10%)

When I trained the student network, is_pi_use, is_pa_use and is_ho_use parameters are False. I used ImageNet pretrined weight except conv1, conv2, conv3 and layer1.0.conv.

How to train for reaching your mIoU?

My environments are Ubuntu 16.04 Pytorch-0.4.1.post2 Python 3.5.6

thanks

irfanICMLL commented 4 years ago

You may get around 68 using the new version of the training code.

irfanICMLL commented 4 years ago

The distillation results will be the same as our paper.

xuyuxiu83 commented 1 month ago

Hi, I trained your script PSPNet-resnet18. (student baseline) But, I can't get refered on paper mIoU(69.10%)

When I trained the student network, is_pi_use, is_pa_use and is_ho_use parameters are False. I used ImageNet pretrined weight except conv1, conv2, conv3 and layer1.0.conv.

How to train for reaching your mIoU?

My environments are Ubuntu 16.04 Pytorch-0.4.1.post2 Python 3.5.6

thanks

hello，I cannot enter the website of the pretrained model now.Could you please share it with me?My email is xuyuxiu83@gmail.com.