FFMetrics
FFMetrics copied to clipboard
fifonik
[Friendly request] Please add source code to the repo :)
Hi,
thanks for providing ffmetrics! It makes these metrics accessible to users like me, who don't know a lot about ffmpeg. I have seen in the TODO.md that you noted about publishing the source.
I would like to endorse this, and opened this issue to show that there indeed users who would love to have the source code available under a license of your choice (personally I would recommend something like MIT).
Even if the sources aren't "pretty" or written for the "public", please consider posting them, noting so in the README and cleaning up sometime down the road if you feel like it, rather than delaying to release until such cleanup has happened.
Please note that this is only my opinion, and I in no way want to sound demanding!
Have a great day and best regards, Tyrius
The author may not wish to publish the source ....due to some genuine reasons....but then a good read me with details on how to make some improvements or modifications on the app should be provided at least....
With time some other video quality metrics may evolve & surpass the video quality efficiency evaluated by VMAF, SSIM or PSMR. for eg. say VQM, MS-SSIM OR BUTTERAUGLI &....there is no way to quickly modify or customize FFmetrics for personal needs.
I think that's the biggest drawback of the FFmetrics ...as stiff competition is being provided by some python apps which shows same output (PSNR, SSIM, VMAF) with better customized graphs.
Any updates on source release? Still very interested in being able to expand the codebase ourselves
Been running into an issue regarding the bad frame extraction functionality, while using this application on my Arch install with Wine. I would have loved to be able and see what the underlying mechanics are so I could maybe have gotten a better idea what breaks on my side. But if open source is not on the table, please say so as well : ). Great tool nonetheless!
Edit: I resolved my issue
This is my 1st C#/WFP project so I'm learning the language and framework. Right now I do not feel comfortable to publish sources -- too many bits in code that I even do not like myself. Sorry, may be sometime in future.
This is my 1st C#/WFP project so I'm learning the language and framework. Right now I do not feel comfortable to publish sources -- too many bits in code that I even do not like myself. Sorry, may be sometime in future.
Thank you for the reply and explanation. I completely understand your reasoning. Release it when you feel comfortable in doing so! But also keep in mind that everyone started at some point ; ) (although, to be fair, some people forget that). Aside from that, open sourcing can also be a way to view how others might solve a problem (if people contribute that is). This can (in my opinion) accelerate one's learning process and broaden the view on how to solve specific problems. Anyways, I think you have a very nice application, thank you for developing and maintaining it. It is a useful tool for me.
I'd also like to add that honest to god this program is already better and simpler than even many commercial tools.
It has been two years since this issue was opened. With some small beta updates earlier in the year that have now slowed down I feel it would be beneficial to open the source to allow external contributions on this project
Bumping, I'm interested in seeing the source code to compare against other implementations VMAF calculation apps. I'm curious how FFMetrics is so fast
You do not need source codes for such analisys. FFMetrics is not calculating VMAF on its own. It is simply calling FFMpeg to do the math. All FFMpeg commands can be saved into log. Feel free to analyze these ffmpeg commands.
I'd personally love to see the source to attempt to add additional metrics such as SSIMULACRA2 or Butteraugli
Understood. I answered to Hunters message related measuring performance. My point was -- FFMpeg doing all the math so its performance should be analyzed if needed.
Sorry I misspoke - I meant how fast the graph generation was. I was generating graphs for some of the project in Python with matplotlib and it wasn't as fast as it is here.