search

HRNet / HRNet-Semantic-Segmentation

The OCR approach is rephrased as Segmentation Transformer: https://arxiv.org/abs/1909.11065. This is an official implementation of semantic segmentation for HRNet. https://arxiv.org/abs/1908.07919
Other
3.12k stars 685 forks source link

The HRNet+OCR branch #102

Open dreamPoet opened 4 years ago

dreamPoet commented 4 years ago

Hi, I just find this branch...is this finished and reliable?

welleast commented 4 years ago

The performance on LIP and Pascal Context (COCO stuff and ADE) is stable. But on Cityscapes, the performance is not very stable.

dreamPoet commented 4 years ago

The performance on LIP and Pascal Context (COCO stuff and ADE) is stable. But on Cityscapes, the performance is not very stable.

Shall I use the same environment setting as the one used for the original HRNet training (I use cuda10, gcc-4.9.4 and pytorch1.2.0)?

Margrate commented 3 years ago

The performance on LIP and Pascal Context (COCO stuff and ADE) is stable. But on Cityscapes, the performance is not very stable.

Can you got mIoU(59 calsses)=54.0 and mIoU(60 classes) = 48.3 ? I can't get the results as paper release.