Hey, how about your test results with pretrained model?

[Lee-CheWei]

I have test the image on GOPRO dataset by running test_matrics.py,
and with pretrained mobilenet model that author release,
but I got PSRN=26.89 and SSIM=0.814, not 28.17 and 0.925.
Does anyone got higher results?

[phoebe1996511]

I found that the link provided by the author was invalid. How to get the pretrained model that author release?

[lisishen]

I found that the link provided by the author was invalid. How to get the pretrained model that author release?

You can find the link in readme.

[Lee-CheWei]

I found that the link provided by the author was invalid. How to get the pretrained model that author release?

Just copy the link to your explorer.

[yff0815]

hello，could you tell me which PyTorch version you install?
when I tested the GoPro dataset by running test_matrics.py,
and with pre-trained Inception-v2 model that author release,
it turns error 'Unexpected key', Maybe it's a Pytorch version mismatch.

[yff0815]

InceptionResNet-v2 pretrained models is fpn_inception.h5, however, the weights parameters is not same with code, do you have same problem.
it turns error 'Unexpected key', Maybe it's a Pytorch version mismatch.
could you tell me which PyTorch version you install?
@Lee-CheWei

[congratulation888]

我已经通过运行test_matrics.py
并使用作者发布的经过预训练的mobilenet模型在GOPRO数据集上测试了图像，
但是我得到PSRN = 26.89和SSIM = 0.814，而不是28.17和0.925。
有没有人获得更高的结果？
Why didn't I change the author code, the d-loss that came out was all negative and different from the author's positive value.

[Lee-CheWei]

InceptionResNet-v2 pretrained models is fpn_inception.h5, however, the weights parameters is not same with code, do you have same problem.
it turns error 'Unexpected key', Maybe it's a Pytorch version mismatch.
could you tell me which PyTorch version you install?
@Lee-CheWei

I don't have that problem, the weights parameters are equal.
Did you change disc_loss in config.yaml ?
And my torch vision is 1.3.0

[congratulation888]

InceptionResNet-v2 pretrained models is fpn_inception.h5, however, the weights parameters is not same with code, do you have same problem.
it turns error 'Unexpected key', Maybe it's a Pytorch version mismatch.
could you tell me which PyTorch version you install?
@Lee-CheWei

I don't have that problem, the weights parameters are equal.
Did you change disc_loss in config.yaml ?
And my torch vision is 1.3.0

How do you allocate the data set, is it the gopro provided by the author? There are 2200 pairs in the train. Do you put the train and valiation paths directly in the config during training? Do other parameters of config need to be adjusted? There is a big gap with the author's results.

[Lee-CheWei]

InceptionResNet-v2 pretrained models is fpn_inception.h5, however, the weights parameters is not same with code, do you have same problem.
it turns error 'Unexpected key', Maybe it's a Pytorch version mismatch.
could you tell me which PyTorch version you install?
@Lee-CheWei

I don't have that problem, the weights parameters are equal.
Did you change disc_loss in config.yaml ?
And my torch vision is 1.3.0

How do you allocate the data set, is it the gopro provided by the author? There are 2200 pairs in the train. Do you put the train and valiation paths directly in the config during training? Do other parameters of config need to be adjusted? There is a big gap with the author's results.

Just do the same thing with #33

[suke27]

there are 2 pretrained models for. is it training using MobileNet?

I have test the image on GOPRO dataset by running test_matrics.py,
and with pretrained mobilenet model that author release,
but I got PSRN=26.89 and SSIM=0.814, not 28.17 and 0.925.
Does anyone got higher results?

do you resolve this problem, meet similar psnr goal with authors?

[hs457681503]

我已经通过运行test_matrics.py
并使用作者发布的经过预训练的mobilenet模型在GOPRO数据集上测试了图像，
但是我得到PSRN = 26.89和SSIM = 0.814，而不是28.17和0.925。
有没有人获得更高的结果？
Why didn't I change the author code, the d-loss that came out was all negative and different from the author's positive value.

你好，我正在复现这个代码，数据集存放的路径应该是怎样的？望回复！

[WongJianyang]

我已经通过运行test_matrics.py
并使用作者发布的经过预训练的mobilenet模型在GOPRO数据集上测试了图像，
但是我得到PSRN = 26.89和SSIM = 0.814，而不是28.17和0.925。
有谁获得更高的成绩？

Hello, after the model training, what files can be run to get the final test results of the author? Is PSNR in training the result?

[WongJianyang]

我已经通过运行test_matrics.py
并使用作者发布的经过预训练的mobilenet模型在GOPRO数据集上测试了图像，
但是我得到PSRN = 26.89和SSIM = 0.814，而不是28.17和0.925。
有谁获得更高的成绩？
你好，模型训练结束后，运行什么文件可以得到作者最终的测试结果？训练时的psnr是结果吗？

[Williamlizl]

Hi, could you tell me the structure of the train sets?

aceback (most recent call last): File "train.py", line 183, in <module> train, val = map(get_dataloader, datasets) File "/home/lbc/DeblurGANv2/dataset.py", line 139, in from_config files_a, files_b = map(list, zip(*data)) ValueError: not enough values to unpack (expected 2, got 0)

[ShiqiLi047]

我使用作者提供的模型fpn_inception.h5 以及 inceptionresnetv2-520b38e4.pth 两个模型进行测试，测试结果在图像上会出现玫红色颜色，请问大家会有这样的现象吗？

[mkt1733]

我使用作者提供的模型fpn_inception.h5 以及 inceptionresnetv2-520b38e4.pth 两个模型进行测试，测试结果在图像上会出现玫红色颜色，请问大家会有这样的现象吗？

小白提问，这个fpn_inception.h5应该放在哪啊

[shuangnan]

Hi, could you tell me the structure of the train sets?

aceback (most recent call last): File "train.py", line 183, in <module> train, val = map(get_dataloader, datasets) File "/home/lbc/DeblurGANv2/dataset.py", line 139, in from_config files_a, files_b = map(list, zip(*data)) ValueError: not enough values to unpack (expected 2, got 0)

could you tell me ，how to fix this problem，thanks