fudan-zvg / gss Goto Github PK

Thanks for putting this code in public. But I'm confused about this empty file, could you upload it at any time?

Could you please upload the posterior learing for x^-1?

Hi~
I try to run tools/pytorch2onnx.py and then return error: " No such file or directory: /home/chenjiaqi/pj/mmsegmentation/ckp/encoder.pkl". So could you share the weight? Thanks!

Thank you for the excellent model you provided. I encountered a problem “FileNotFoundError: EncoderDecoder: GenerativeSegHeadFTW: [Errno 2] No such file or directory: 'ckp\encoder.pkl'” when I tried to reproduce it. How should I solve it?

Excellent research! I was wondering if you could add a LICENSE file to your repo? Thank you!

首先感谢作者公开这一出彩的工作，但貌似比较核心的代码尚未公布，如果作者能公布Latent prior learning和Latent posterior learning 这部分的代码就太好了！

.

我使用提供的基于cityscapes的FT-W-Swin-L进行了验证，结果只有0.0011，请问这个checkpoint提供的准确嘛，因为在cityscapes下其他三个我都得到了报告的结果，只有这个特殊。

我在用你们的预训练权重去推理cityscapes测试集的时候，sample per gpu设置为1，10G的显存还不够用，请问是可能是哪里出了问题？还是单纯的你们的模型需要很大的显存呢？（之前没有遇到过需要这么多显存的模型）

Thanks for the nice work.

However, some issues remain if we want to train this model from the stretch on one's own dataset:

• Is the stage-1 training code and Maskige reconstruction code available now?

• Is there any instruction on how to train the model? Now there is no CMD sample in this webpage.

Thanks!

如果一段文字的描述有對應一張圖, 那用你這個架構然後在學posterior的時候把總結文字的latent code跟影像的latent code做matching是不是就可以跟據文字生成影像了? 甚至聲音也可以?