z-bingo/fastdvdnet

An unoffical implement of FastDVDNet by PyTorch

• 👋 Hi, I’m @z-bingo
• 👀 I’m interested in computer vision especially its applications on autonomous driving

fastdvdnet's Issues

You are missing `nn.PixelShuffle` Layer

Description

Well, the denoising results seems good but you may wonder why the artefact in you results?
In contrast to the original paper implementation of original author (github), you simply did not have nn.PixelShuffle.

What does nn.PixelShuffle do you ask? In short it which helps reducing gridding artefacts stated by Krizhevsky et al..

I can push a fix but with original author (github)
already available and accurate, I won't bother to push. But it's good in case you didn't know

Hi, z-bingo

Hi, first of all, thank you for sharing your codes.

I have run them and I have an issue.

I have found some artifacts in the result images as you have mentioned.

Have you found out why these artifacts are occurred?

and can you tell me the average evaluation PSNR (with vimeo 90k) please?

It would be great if you give me some help.

thanks.

About psnr results

Hi, z-bingo, have you compared the psnr results between the model you trained and the model provided by author? As I see lots of artifacts in your results.

con4 and con3 feature map numbers ERRORS

Hi, don't you think your setting of feature maps number have some problems?

请问您左上角绿点的问题解决了吗？

最近几天用您的代码在我的数据集上训练，似乎也出现了您说的“for example, the left side of this image has a large area of green spots.”这个问题，（加载不出来您的结果，推断应该是同一个问题）请问这个应该是由什么造成的？

About dataloader ERROR

C:\Users\Win10\Desktop\unet-master\venv\Scripts\python.exe C:/Users/Win10/Desktop/unet-master/FastDVD-master/train_eval.py
Namespace(batch_size=64, cuda=False, dataset_path='F:/JPY/vimeo_triplet/sequences', eval=False, frames=2, im_size=96, learning_rate=0.0001, max_epoch=100, num_worker=4, restart=False, txt_path='F:/JPY/vimeo_triplet')
There is no any model to load, restart the train process.
Traceback (most recent call last):
File "C:/Users/Win10/Desktop/unet-master/FastDVD-master/train_eval.py", line 200, in
train(args)
File "C:/Users/Win10/Desktop/unet-master/FastDVD-master/train_eval.py", line 92, in train
for iter, (data, gt) in enumerate(data_loader):
File "C:\Users\Win10\Desktop\unet-master\venv\lib\site-packages\torch\utils\data\dataloader.py", line 521, in next
data = self._next_data()
File "C:\Users\Win10\Desktop\unet-master\venv\lib\site-packages\torch\utils\data\dataloader.py", line 1203, in _next_data
return self._process_data(data)
File "C:\Users\Win10\Desktop\unet-master\venv\lib\site-packages\torch\utils\data\dataloader.py", line 1229, in _process_data
data.reraise()
File "C:\Users\Win10\Desktop\unet-master\venv\lib\site-packages\torch_utils.py", line 425, in reraise
raise self.exc_type(msg)
FileNotFoundError: Caught FileNotFoundError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "C:\Users\Win10\Desktop\unet-master\venv\lib\site-packages\torch\utils\data_utils\worker.py", line 287, in _worker_loop
data = fetcher.fetch(index)
File "C:\Users\Win10\Desktop\unet-master\venv\lib\site-packages\torch\utils\data_utils\fetch.py", line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "C:\Users\Win10\Desktop\unet-master\venv\lib\site-packages\torch\utils\data_utils\fetch.py", line 44, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "C:\Users\Win10\Desktop\unet-master\FastDVD-master\data_provider.py", line 50, in getitem
img = Image.open(file)
File "C:\Users\Win10\Desktop\unet-master\venv\lib\site-packages\PIL\Image.py", line 2912, in open
fp = builtins.open(filename, "rb")
FileNotFoundError: [Errno 2] No such file or directory: 'F:/JPY/vimeo_triplet/sequences\00068/0821\im4.png'

Process finished with exit code 1

A question about test process

excuse me, can you tell how to denoise my own video?I am a little confused, thanks