Generative_Adversarial_Perturbations icon indicating copy to clipboard operation
Generative_Adversarial_Perturbations copied to clipboard

Published 20 hours ago • verified publisher icon OmidPoursaeed

FCN-8s trained to segment Cityscapes

Open CNHNLP opened this issue 4 years ago • 3 comments

hello,I use the pretrained model (FCN-8s trained to segment Cityscapes) to evaluate on the cityscapes val set, and the IOU is very poor (44%). Have you confirmed the pretrained models?

CNHNLP avatar Mar 12 '20 08:03 CNHNLP

Hi, I'm sorry for disturbing you, because of the invalid link, could you please share "fcn-8s-cityscapes" weight to me? thank you!

[email protected]

kravrolens avatar Jul 13 '21 02:07 kravrolens

Please consider this: https://github.com/wkentaro/pytorch-fcn

OmidPoursaeed avatar Jul 13 '21 06:07 OmidPoursaeed

Please consider this: https://github.com/wkentaro/pytorch-fcn

Hi, I downloaded the pretrained model. But the tar file cannot be extracted and recognized. And the link of pytorch-fcn only provides model trained on pascalvoc. Is there any pretrained models for cityscapes dataset?

Besides, I tried to train a segmentation model. But it requires a 'imgsets.txt' for each file. Is the official dataset not the one you used? @OmidPoursaeed

ghost avatar Jan 17 '22 13:01 ghost