Check motion magnification

[nicolas-tardieu]

Hi all,

First, thank you for the work done since I am deeply interested in the potential of this algorithm and mainly in motion magnification.
I managed to compile the code but I am still strugling to be sure it works OK.

Would you have an example video and dedicated settings to check it works fine on my computer?

Thanks,
Nicolas

[tschnz]

Hey, I can verify that magnification with Riesz Pyramids does NOT work as expected due to a bug.

[nicolas-tardieu]

Thanks for the information Jens.
Do you plan to fix the bug or you do not have time for that at the moment?

[nincube8]

You have to change the lines in the pro file to match the directories where your files are located. And you need to install the prerequisites until you have all the files you need. I recommend using MobaXterminal to execute if you are in windows 10. "sudo qtcreator rvm.pro" will open the program in QT so that you can compile and run it. If you haven't met the prerequisites, go back to square one and start again. The prereqs need to be installed correctly, and it's not the easiest.

…

On Fri, Sep 25, 2020 at 12:45 AM nicolas-tardieu ***@***.***> wrote: Thanks for the information Jens. Do you plan to fix the bug or you do not have time for that at the moment? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEBLRRKF3OY5K2A7K7KNAOTSHQ4CLANCNFSM4RYNNSNQ> .

[nincube8]

The rvm.pro file will not have lines to directories that will match yours by default, you need to CORRECTLY redirect to the correct directories from the rvm.pro file. On Fri, Sep 25, 2020 at 1:27 AM CK Gaming Channel <[email protected]> wrote:

…

You have to change the lines in the pro file to match the directories where your files are located. And you need to install the prerequisites until you have all the files you need. I recommend using MobaXterminal to execute if you are in windows 10. "sudo qtcreator rvm.pro" will open the program in QT so that you can compile and run it. If you haven't met the prerequisites, go back to square one and start again. The prereqs need to be installed correctly, and it's not the easiest. On Fri, Sep 25, 2020 at 12:45 AM nicolas-tardieu ***@***.***> wrote: > Thanks for the information Jens. > Do you plan to fix the bug or you do not have time for that at the moment? > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#10 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AEBLRRKF3OY5K2A7K7KNAOTSHQ4CLANCNFSM4RYNNSNQ> > . >

[nicolas-tardieu]

Perhaps I was not clear enough : I did compile the code and I can run it on a video.
My question is : can you provide me with a test video so that I can be sure my installation is fine as far as motion magnification is concerned (usually, there is a target "test" in a Makefile that runs nonregression checks).

[nincube8]

When you run the program, it gives you an option to open a video file, or a webcam interface. Then you can use the sliders and preview the effects in real time. You should see it working, if not, then maybe one of the directories is off, or you build opencv incorrectly, which I did a couple of times. It took me many tries to get a successful run. However, the program is not completely stable, and has bugs, and will crash many times. It's still a cool program though. Thanks to tschnz.

…

On Fri, Sep 25, 2020 at 2:15 AM nicolas-tardieu ***@***.***> wrote: Perhaps I was not clear enough : I did compile the code and I can run it on a video. My question is : can you provide me with a test video so that I can be sure my installation is fine as far as motion magnification is concerned (usually, there is a target "test" in a Makefile that runs nonregression checks). — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEBLRRO3KII2B7B3KIFLFCLSHRGRTANCNFSM4RYNNSNQ> .

[MirekJan]

I suggest you grab the original data from the paper found here: http://people.csail.mit.edu/nwadhwa/riesz-pyramid/ Happy motion magnifying 🙂

…

On Fri, Sep 25, 2020 at 9:15 AM nicolas-tardieu ***@***.***> wrote: Perhaps I was not clear enough : I did compile the code and I can run it on a video. My question is : can you provide me with a test video so that I can be sure my installation is fine as far as motion magnification is concerned (usually, there is a target "test" in a Makefile that runs nonregression checks). — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABUYX7OTCCSLIXZBFETFUGDSHRGRVANCNFSM4RYNNSNQ> .

[tschnz]

This was my bachelor thesis and I won't find (unpaid) time to add tests and check for validity. The code makes me shiver already but I try to keep it compileable. The code is not correct in terms of all framerates and wavelength, especially with varying FPS on live streams but it serves as a good enough proof-of-concept and to showcase that live amplification is possible. I would not use this code in a business context at all (I tried once). It's error prone and best case needs a shader based rewrite.

I leave this issue open as warning to other users.