Longguang Wang

Results 89 comments of Longguang Wang

Hi @cunese0666 , thanks for your interests in our work and sorry for the late response. 1. When the batchsize is set to 1, please modify the code `net =...

Hi @huohuohuohuohuohuohuohuo,你可以使用finetune.py在KITTI数据集上进行训练,也可以在train.py中将`SceneFlowDataseet`替换为`KITTIDataset`进行训练。

Hi @huohuohuohuohuohuohuohuo,运行是可以运行的,只不过会慢很多。如果是训练的话,建议还是在GPU上进行;如果是想在cpu上测试的话,请将这里的[device](https://github.com/LongguangWang/PAM/blob/master/PASMnet/test.py#L14)修改成cpu。

Thank you all for your interests in our work. I've been busy with a recent DDL and I'll release the code after that in late Nov.

Hi @zyl1336110861 @qingxi2020 @MJITG @xie1234 @LihengXu, the code is made available now.

是的,是一维的。送入超分网络预测kernel前没有reshape(https://github.com/The-Learning-And-Vision-Atelier-LAVA/DASR/blob/main/model/blindsr.py#L37),预测调制系数时有进行reshape(https://github.com/The-Learning-And-Vision-Atelier-LAVA/DASR/blob/main/model/blindsr.py#L62)

您好,您理解的没错,我们将同一LR图像中提取的其他图像块视为正样本,来自其他LR图像的图像块视为负样本。 具体到代码里也是这么操作的,dataloader里完成了的图像块的裁取,所以送入网络的就已经是图像块了。您可以再具体一下疑惑的代码位置方便进一步解释您的疑惑。

感谢关注~self.queue中保存了之前iteration送入的patch样本。由于不同图像之间的退化是不一样的,因此这些之前iteration送入的、保存在self.queue中的样本对于当前的query来说就成了负样本。

Thanks for your interests in our work. Q1: `table_q = table_q.data + (table_q - table_q.data) / wgt * x.numel() / (table_q.numel() // 2 + 1)` still preserves the gradients of...