Kento Doi
Kento Doi
Did you solve it? I have the same problem.
@mees I solved the problem by commenting out lines 39,58,59,78,79,80,81. Maybe the calculation of the loss is not needed.
Could you reproduce the score of the original Densecap parper? I have the same problem and consider to train a new model from scratch.
I got same error in my macbook. It seems to be the error because of the out of memory...
No, I don't. If you can, could you add CRF, please?
Thank for your comment. I haven't tried yet. I'm sorry that I think it's take a long time to experiment with Pascal or Cityscapes dataset.
Hello. It may be my mistake. If you'd like to, please send me pull request. Thank you.
Which dataset do I use? If you have implementation of dataset, please share for me.
It is ResNet pretrained with ImageNet. (officially provided by PyTorch community) Actually, the official implementation of DeepLab is available, maybe it is better than my implementation.
Please see https://github.com/sxyu/svox2/issues/11 You can find code to create a mesh and It may help you.