Results 6 comments of wangqiim

> Hi @btwbtm, thanks for your interest in our work. Softmax is also used in several network quantization or pruning methods to soften one-hot distributions. In my opinion, softmax may...

> 你好,请问你把class loss 和 average loss应用到降噪上最后work了吗?我发现数据集如果不划分(即初始branchs都在一个数据集上train),最后classifier不会收敛= =

> > > > > > > > > 你好,请问你把class loss 和 average loss应用到降噪上最后work了吗?我发现数据集如果不划分(即初始branchs都在一个数据集上train),最后classifier不会收敛= = > > work的,文章接近最后有放DNCNN的结果,但我觉得如果用到去噪上还可以做一下patch大小之类的适配(因为超分会x4,去噪相当于只x1) 对,这样不会收敛的。初始branch放在一个数据集上train会导致他们没有各自“擅长”的部分,这个知识也就没法在联合训练里传递给classifier了。 明白了!非常感谢,这个classfier的训练看起来容易操作起来不容易呀,我尝试改了改都会崩= =

Yes, in this code it can't improve speed, because it excatly excute all conv(need be skip by test if times 0), then times 1 or 0, it represent excute or...

I switch different resolution use ```python input # [batchsize, 3, 32, 32] inputs = [ F.interpolate(input, size=[32, 32]).cuda(), F.interpolate(input, size=[16, 16]).cuda(), F.interpolate(input, size=[8, 8]).cuda() ] ``` instead of modify dataloader....

> > In Fig.4.(a), I find different models will cause different PSNR lookup results. How to get Spearman correlation (0.85) coefficient picture (Fig.4.(a)). What model do you use to generate...