mrgreen3325
mrgreen3325
> You may need to set the parameters begin_norm_axis=-1 > nn = Conv2d(64, (3, 3), (1, 1), padding='SAME', W_init=w_init, b_init=None)(n) > nn = LayerNorm(begin_norm_axis=-1, act=tf.nn.relu)(nn) Thanks laicheng. May I know...
> As indicated by the name, `prepare_testset.yml` is used for generating testing set. For training, the HR-LR pairs are generated one-the-fly, which means only HR path is required. Thanks. Yes,...
> We generate HR-LR pairs on-the-fly. The parameters on degradation are set at > > https://github.com/JingyunLiang/MANet/blob/34f90ba8888f4a1dd2a1127b97c2ec3706f06598/codes/options/train/train_stage1.yml#L10 Thanks for reply. I follow the train_stage1.yml setting to config prepare_testset.yml to process 4x...
> Awesome! Thanks for clarifying my questions! > > One additional comment, I don't know if you had seen this work before, but it uses SRMD and USRNet (among others)...
> hello, i using google collab Tesla T4gpu memory of 15 gb and takes 7k training images for 50 epochs but after 30 epochs it always says that cuda out...
> 问题3： 效果测试？ 我直接下载 NTIRE2020 中的 [测试数据，test_track](https://data.vision.ee.ethz.ch/alugmayr/real-world-sr/NTIRE20/track2-test-input.zip)，经过您发布的[模型](https://pan.baidu.com/s/1BcYcX0yCS-3-6XqT4BgYAQ?pwd=ovmw)测试数据不知为何发现效果贼拉差。 示例如下： > > 原图 作者的图 bsrgan 基于自己的数据集模型(未用到该数据) >     >     > 取的数据集的首尾共两张。发现您这个模型的效果稍稍微微有点差啊。 请问您这篇论文的主题主导为您提出的这种退化思路，而不重在实际的模型超分效果吗？ Hello, 基于自己的数据集模型, are...
> （1）问题一：KeyError: 'netDeg' Traceback (most recent call last): File "train.py", line 481, in main() File "train.py", line 171, in main main_worker(0, 1, opt, args) File "train.py", line 305, in main_worker...
> I have tried my best to train 2020Track1 by using psnr/2020Track1.yml and pretrained SR weight edsr_baseline_x4-new.pt, but I just can get PSNR 26.2 and cannot get the same PSNR...
> After I changed '**the loss weight for LR guidance(from 4 to 40)**' and continually trained the pretrained 'IRN_x2' model with 5000 epochs, the model is more robust with all...
> > > After I changed '**the loss weight for LR guidance(from 4 to 40)**' and continually trained the pretrained 'IRN_x2' model with 5000 epochs, the model is more robust...