junyaohu / common_metrics_on_video_quality Goto Github PK
View Code? Open in Web Editor NEWYou can easily calculate FVD, PSNR, SSIM, LPIPS for evaluating the quality of generated or predicted videos.
You can easily calculate FVD, PSNR, SSIM, LPIPS for evaluating the quality of generated or predicted videos.
What's the difference between i3d_pretrained_400.pt and i3d_torchscript.pt? I don't know much about this.
How do I use it when I test fvd?
How can I read a mp4 file from a path, not directly generate a tensor, can you give me a implementation code
"Thank you for your work, I have a question about clip_timestamp. I noticed that when calculating the FVD (Fréchet Video Distance) metric, only the first clip_timestamp frames are taken into account. Is this a reasonable approach? Does it meet the requirements of the paper? For example, if I want to test related metrics on the UCF101 dataset, how should I choose?"
Hi, thanks for your codebase! In calculate_lpips.py
, it says image should be RGB, IMPORTANT: normalized to [-1,1]
. https://github.com/JunyaoHu/common_metrics_on_video_quality/blob/c12e449540cf27c1b1412c8012ecc35413fda6a7/calculate_lpips.py#L26C59-L26C59 However, in the README notice
section, it says Make sure the pixel value of videos should be in [0, 1].
common_metrics_on_video_quality/README.md
Line 167 in c12e449
Hi! Thanks for your work. I have this error when running demo.py: RuntimeError: forward() is missing value for argument 'x'. Declaration: forward(torch.src.i3dpt_jit.I3D self, Tensor x, bool rescale=False, bool resize=False, bool return_features=False) -> Tensor. It comes from line 27 in fvd/styleganv/fvd.py. It seems that the input is taken as self' and no value for argument
x'.
why fvd resize img to (224, 224), Is it suitable for high-resolution?
Nice practical tool kit !
I have a question about using the code to calculate FVD on a whole dataset. I have thousands of videos, and I think it is hard for me to calculate FVD simply by setting the parameter 'NUMBER_OF_VIDEOS' to e.g. 4000 considering the computation cost.
So, how should I calculate FVD on a large dataset, maybe by iteratively calculating batch by batch and then average the number ?
Thanks in advance!
I have a question. When testing indicators, should the number of frames and the number of videos generated by the algorithm be the same as the test video?
Hello. In the README you say that with scipy>1.9 FVD calculations will be wrong. Could you, please, fix the calculation method for newer versions of scipy or at least elaborate further on why that happens?
Hello, how can I calculate the FVD of 8 frames short video? Thank you!
hi,
how come you calculate the fvd score with the output of the logits layer? Doesnt the idea of the FID and FVD come from using the feature space?
0
to numframe-1
.
videos_clip1 = videos1[:, :, : clip_timestamp + 1]
videos_clip2 = videos2[:, :, : clip_timestamp + 1]
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.