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
3.16k stars 690 forks source link

have you train pascal voc 2012 dataset? #24

Open GuangyanZhang opened 5 years ago

GuangyanZhang commented 5 years ago

I had train the pascal voc 2012 dataset,but The result of training is 67.23, the training loss is decreaseing, but the valing loss unchange,Can you tell me your result?thank you.

sunke123 commented 5 years ago

We have not trained HRNets on the pascal voc dataset.

YijianLiu commented 5 years ago

Have you met this problem: ‘RuntimeError: weight tensor should be defined either for all or no classes at /opt/conda/conda-bld/pytorch_1535491974311/work/aten/src/THCUNN/generic/SpatialClassNLLCriterion.cu:27’

dawnstar2 commented 4 years ago

I had train the pascal voc 2012 dataset,but The result of training is 67.23, the training loss is decreaseing, but the valing loss unchange,Can you tell me your result?thank you.

I met the same problem. I trained several times, all the scores are around 67.0. I guess my configurations were improper, but I have no idea how to modify the configure file.