tencentarc / animesr Goto Github PK

Hello, I followed your codes, but got error randomly in animesr.data.ffmpeg_anime_dataset.FFMPEGAnimeDataset.add_ffmpeg_conpression. I may get "[h264 @ 0x1d1df20] decode_slice_header error" in terminal, when the program runs to "ffmpeg_video2img.stdin.close()".
In this case, 15 images are passed in to generate video, but decoding can result in 14 or fewer images, causing problems. I don't know why this problem occurs, and what's bothers me most is that it is arisen randomly. Do you have any suggestions or ideas?

Hey guys,

i encountered a big while installing where the program tells me that the Model was not found in the weight Folder.
But it is in the weight Folder, already tried reinstalling and so on.

Above the console log.

Thanks for any help in advance!

Hi,

In your paper, you mention that you fine-tuned BSRGAN, Real-ESRGAN, and RealBasicVSR with your dataset and shared the visual results of these fine-tuned models. Do you plan to share the fine-tuned weights of these models as well for comparison purposes?

Thanks.

Hello! Thanks for your great contribution to the Anime restoration domain. AnimeSR is a very interesting paper and the performance is really good. I wonder if there is any weight provided for fine-tuned Real-ESRGAN, BSRGAN, and Real-BasicVSR told in the paper? Thanks!

Hi Guys, it's me again, i really started liking your project and was questioning myself if you guys want to keep working/updating it for the future since i have not found any roadmaps or information about that.

Really love it keep it going!

What can cause this error when running a train?

load_net = load_net[param_key]
KeyError: 'params'
    self.load_network(self.net_g, load_path, self.opt['path'].get('strict_load_g', True), param_key)    load_net = load_net[param_key]
  File "/mnt/ai_data/jw_chae/tmp/Super_Resolution/AnimeSR/env/lib/python3.8/site-packages/basicsr/models/base_model.py", line 295, in load_network

Hello! I'm glad to get the AVC dataset. However, I found thousands of pictures which are corrupted in AVC-Train. At first, I thought my PC has something wrong. So, I download the AVC dataset three times. But the same problem occurred every times. Is there bug in AVC-Train dataset?

"If you are looking for portable executable files, you can try our realesr-animevideov3 model which shares the similar technology with AnimeSR."

Could you release the training config of realesr-animevideov3. I'm very interested in the difference between realesr-animevideov3 and the previous models.

Thanks.

Hi, Thanks for your wonderful work.
In https://github.com/TencentARC/AnimeSR/blob/main/animesr/data/ffmpeg_anime_dataset.py
line 103: img_bytes = self.file_client.get(img_gt_path, 'gt') and
line 130: ffmpeg.input('pipe:', format='rawvideo', pix_fmt='rgb24', s=f'{width}x{height}',
I have check the code, line 103 will read a image with 'bgr', but in line 130, pix_fmt='rgb24'.
Is this a mistake？

I'm looking for good approaches in super video resolution, and I found out your proposal quite interesting. I think that the way you train your model to learn a degration space is really effective.

So, I was wondering if that idea would work for real videos. I have a 4k 60fps video dataset, and I want to know if using would be enough for train a new model.

Hello, I have read your work and was deeply inspired by it, really great work! Would you please release the code & dataset sooner in the feature?
Thanks a lot!

The multiprocessing function of the code does not initiate unless there is an Nvidia graphics card available. rerouting to CPU it looks like there isn't an algorithm for the CPU. Is this to be expected?

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.

If you are interested in participating, you can add your algorithm following the submission steps:

• Submit for Video Upscalers Benchmark: Quality Enhancement
• Submit for Super-Resolution for Video Compression benchmark

We would be grateful for your feedback on our work!

When I do step 1 of training, where can I modify the settings used for the classic basic operators (blur, noise, etc)?
Does this go in the options yaml file? Or do I need to modify one of the python files?

Hi,

Thanks for providing this amazing projects. I tried the learning step 1, which trained 300k iterations with half of the provided AVC-Train datasets. Every setting in the train_animesr_step1_net_BasicOPonly.yml file was kept exactly the same. However, it appears to me the provided pretrained_animesr_step1_net_model.pth achieved much better performance than the model I trained (the 300k iteration is not done yet, I tried net_g_170000.pth). Would you mind shed some lights on the possible mistakes I might took. Thank you.