|
From: | Aditya Arun Kumar |
Subject: | Re: using gr-specest/gr-inspector |
Date: | Fri, 14 May 2021 16:52:22 +0530 |
Sender notified by Mailtrack 05/14/21, 04:51:09 PM |
Hi Aditya!
I think the most straightforward way would be to port the AMC block to 3.8. I wasn't aware
of it existing only in maint-3.7!
Best regards,
Marcus
On 14.05.21 00:50, Aditya Arun Kumar wrote:
> Hi Marcus,
> Thanks for the reply. The reason I am sticking with GRC 3.7 branch is that the AMC block
> is present only in the gr-inspector 3.7 branch, which requires a CNN graph for input, I
> have no issues in running the 3.8 branches of the OOT. So, if this is the case is there a
> way I can import a TensorFlow graph model into GRC3.8 because the inspector branch really
> says it is not present.
>
>
>
> Mailtrack
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&>
> Sender notified by
> Mailtrack
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&>
> 05/14/21, 04:14:53 AM
>
>
> On Fri, May 14, 2021 at 4:08 AM Marcus Müller <mmueller@gnuradio.org
> <mailto:mmueller@gnuradio.org>> wrote:
>
> Hi Aditya,
>
> you shouldn't use GNU Radio 3.7 for new things in 2021, it's a dead end; we don't develop
> this any further, and it will not work on newer machines; there's nothing we can do about
> that.
>
> Both gr-inspector as well as gr-specest work with GNU Radio 3.8, if you check out their
> maint-3.8 branch.
>
> GNU Radio 3.8 is easily installable on any reasonably modern operating system, so that's
> what you should do.
>
> Neither of this has anything to do with Neural Networks, so you're on your own there.
>
> Best regards,
> Marcus
>
> On 13.05.21 23:38, Aditya Arun Kumar wrote:
> > Hello,
> > I am working on AMC(Automatic Modulation Classification) using GRC. After some digging
> > around I found that gr-inspector along with gr-specest will work for this. I also found
> > that while using gr-inspector I need to check out the maint-3.7 branch (gr-inspector for
> > maint3.7 says that a minimum of GRC 3.7.9.3 is required) for me to load the tf
> models. Now
> > I face problems in 2 places,
> >
> >
> > 1. gr-specest is not really available for 3.7 at least in the kit-cell branch, but
> > available here <https://github.com/robertwatsonbath/gr-specest-3.7
> <https://github.com/robertwatsonbath/gr-specest-3.7>>. So I install that and
> > try to run the example flowgraph I get "Module not found: _swig_specest", the same
> > happens when I try to do it from the terminal as well. After this, I found this
> > <https://signalsintelligence.wordpress.com/
> <https://signalsintelligence.wordpress.com/>> blog that talks about the problems I have,
> > but the problem is I cannot find any of the links in git-repo as most of them are dead.
> > Now my question in this place is there a specific gr-specest branch for GR3.7.
> > 2. The second part of the problem is where can I find a pre-trained CNN model that I can
> > import and use (I am still learning the NN parts). If I cannot find is there a
> > well-documented t project that I can use to generate my CNN graph that I can import
> as AMC
> > into gr-inspector tensor-flow block.
> >
> > But whatever it looks like I am going to need a CNN for mod rec at the moment, but what
> > puzzles me is I cannot find gr-specest anywhere just to understand some 3D visualization
> > as well for 3.7. Any help in this aspect will be really helpful.
> >
> > --
> > S. Aditya Arun Kumar
> >
> >
> > Mailtrack
> >
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&>>
> > Sender notified by
> > Mailtrack
> >
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&>>
> > 05/14/21, 02:54:51 AM
> >
>
>
>
> --
> S. Aditya Arun Kumar
> +919123517465
[Prev in Thread] | Current Thread | [Next in Thread] |