Hi @ckkelvinchan , I have tried to run the denoise pretrianed model with DAVIS-test dataset, but the result images has no difference between the input images. My experiment is as below: pretrained model: basicvsr_plusplus_denoise-28f6920c.pth optical flow model: I download it to my local path dataset: I use the the tractor folder in DAIVS-test dataset which is used in your paper. I add the sigma 50 gaussian noise to the pictures as input. result:

It seem that it run successfully, but the result has no change between the input. Is their any parameter or option error in my attempt? By the way, basicvsr_plusplus_c64n7_8x1_600k_reds4.pth can run successfully and the result is excellent!

why the cond_n1 is the out of flow_wrap(feat_prop, flow_n1.permute(0, 2, 3, 1)) while feat_prop is zeros? why not flow_wrap(feat_n1,flow_n1.permute(0, 2, 3, 1) and feat_n1 = feats[module_name][-1]

I installed pytorch using the following command conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch I am trying to install mmcv-full. I am facing the error

Hello,

MSU Graphics & Media Lab Video Group has recently launched two new Super-Resolution Benchmarks.

• Video Upscalers Benchmark: Quality Enhancement determines the best upscaling methods for increasing video resolution and improving visual quality.
• Super-Resolution for Video Compression benchmark aims to test Super-Resolution methods on compressed videos and select the best model for each video codec standard.

Your method achieved 22nd place in Video Upscalers Benchmark: Quality Enhancement in 'Animation 2x' category. We look forward to your future work!

We would be grateful for your feedback on our work.

Hi, thank you very much for your excellent work. I wonder could you please show us the folders of the dataset like the following forms? So that I can recollect the dataset into the same form as yours.

First of all, thanks for your great work. I have a question for a piece of code which is about flow based DCN

offset_1 = offset_1 + flow_1.flip(1).repeat(1, offset_1.size(1) // 2, 1, 1)

Could you explain why flip(1) is needed for flow, that is why need to switch x and y

In the denoising task, what are the training data set, verification data set and test data set respectively, can you provide them? Is the Set8 dataset a validation set in the code? But in Table5 and Table6 of the paper, we see that Set8 is used for testing

In the technical report for deblurring and denoising (https://arxiv.org/pdf/2204.05308.pdf), there are two versions of the BasicVSR++ for video denoising, the one which reduces the resolution by 2 (BasicVSR++2, superior performance but less speed) and by 4 (BasicVSR++4, more speed but worse performance). Current code, as I understand, implements the version which reduces the resolution by 4. How do I change the code exactly to switch to the BasicVSR++2 (the one which reduces the resolution by 2)? Or could you please share the code?

In addition, I am trying to speed up the training for denoising. I tried increasing samples_per_gpu config variable from 1 to 3 and decreasing the number of iterations from 600,000 to 200,000 (so that the amount of data that the model uses for training is the same), which helped to improve the time quite well (from 13 days to 6 days of training). Is there any other way to make the training faster without hurting the performance after the training?

Thank you!

My Enviorment: PyTorch 1.11.0 Python 3.8(ubuntu20.04) Cuda 11.3 GPU:RTX A5000 * 1;video memory:24GB CPU:14core Intel(R) Xeon(R) Gold 6330 CPU @ 2.00GHz ram:30GB system disk:20 GB data disk:50GB SSD While running the demo based on the basicvsr_plusplus_reds4.pth checkpoint, I put in a blurry video of myself for several tens of seconds, and after a while the process was killed. I tried to set -- max-seq-len=1 but still the process was killed.How can I solve this problem?

When I try to run the training for denoising using the suggested script and configuration file, I get "ValueError: The input sequence is not long enough to support the current choice of [interval] or [num_input_frames]" at BasicVSR_PlusPlus/mmedit/datasets/pipelines/augmentations.py, line 1053. I investigated the code a bit and it seems that the problem is with the way data is loaded. It looks like there is a variable at some point of loading the data which is called sequence_length and it is always zero because it calculates number of png images. However, DAVIS dataset only consists of jpg images, so that's why the sequence_length is zero. I assume the data is preprocessed for denoising task in some way and saved in png format in the end. The data preprocessing step is unfortunately missing in this repo, so I am wondering could it be the reason for my error? Or there is no preprocessing when training for denoising? I am trying to run the training with the DAVIS dataset.

Curious if you have any clue why my error is occurring.

Thank you!