[IPOL discuss] A library of python bindings for IPOL algorithms
martin de la gorce
martin.delagorce at gmail.com
Fri May 6 15:04:19 CEST 2016
Hi all,
It would be nice to have the project advertised on the main page once it
will be good enough (more papers and tested on various os and python
versions). yes , by default one can not just download the wrapper for a
single code but for the whole bundle. However I added in the readme
an explanation on how to install only a subset of the papers.
As more papers will be added, i wonder if installing all the papers will
become a problem in term of the duration of the installation and the size
of the library on the disc. It seems that the code of a paper takes on
average 2Mo after compilation and there are about a hundred papers on IPOL
which makes 200Mo which is quite big for a python library. One solution for
the size problem could be to specify for each paper the list of files or
file extensions than need to be kept after the compilation. Which would
lead to an average of 500ko per paper. For the compilation time, maybe we
could provide prebuild packages through PyPI (https://pypi.python.org/pypi).
about the time needed to add a pape i think i was a bit too optimistic , it
can take almost 2h for some papers.
Best
Martin
On Thu, 5 May 2016 at 11:28 <discuss-request at list.ipol.im> wrote:
> Send discuss mailing list submissions to
> discuss at list.ipol.im
>
> To subscribe or unsubscribe via the World Wide Web, visit
> https://tools.ipol.im/mailman/listinfo/discuss
> or, via email, send a message with subject or body 'help' to
> discuss-request at list.ipol.im
>
> You can reach the person managing the list at
> discuss-owner at list.ipol.im
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of discuss digest..."
>
>
> Today's Topics:
>
> 1. Re: A library of python bindings for IPOL algorithms
> (Jean-Michel Morel)
> 2. Re: A library of python bindings for IPOL algorithms
> (Miguel Colom)
> 3. Re: A library of python bindings for IPOL algorithms
> (Jos? Luis Lisani)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Thu, 5 May 2016 10:01:42 +0200
> From: Jean-Michel Morel <moreljeanmichel at gmail.com>
> To: general discussions about the IPOL project <discuss at list.ipol.im>
> Subject: Re: [IPOL discuss] A library of python bindings for IPOL
> algorithms
> Message-ID:
> <CALpGdHcGCNgpcLuBCDiLxf=
> FMtW0sPKRmYBqG9wcUdLxORxsGw at mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
>
> Dear Martin, and all,
> This seems to me very important as many potential users of the algorithms
> do not use C++.
> Thus I would suggest to Jos? Luis and Miguel to consider including the link
> to the Python interface in the web page of each paper for which it is
> ready.
> Miguel and Jos? Luis, is that possible?
> Best,
> Jean-Michel
>
>
> Jean-Michel Morel
> moreljeanmichel at gmail.com
>
> On Wed, May 4, 2016 at 11:11 AM, martin de la gorce <
> martin.delagorce at gmail.com> wrote:
>
> > Hello
> >
> > I have started a project on Github (
> > https://github.com/martinResearch/PyIPOL) that aims at providing a
> python
> > interface to each of the algorithms that are on IPOL with an example for
> > each algorithm.
> >
> > I think that having Python bindings for the IPOL algorithms can be
> > interesting to people who want to test easily and integrate quickly
> these
> > codes in their applications written in python or in order to script tests
> > easily . The goal is to put it on the official python packages index
> > website (https://pypi.python.org/pypi) which could potentially increase
> > the visibility of the IPOL algorithms.
> >
> > I have for now interfaced about a dozen of papers and the goal is to get
> > almost all of the IPOL papers interfaced with python. Most of the
> > interfaces have been done by calling the original executable with
> temporary
> > files. Some bindings have been written using Cython.
> >
> > Adding a paper takes between 15 min and an hour, depending on the
> > complexity of the input/output formats for the executable. Writing the
> code
> > to get nice drawings of the results in python might take a bit of time
> when
> > there is not yet another paper that perform a similar task already
> > interfaced on whose code can be reused (drawing point matches for images
> of
> > different sizes for example, or optical flow vector fields).
> >
> > I invite you to install a test my bindings and to make suggestions for
> > improvements and contribute by adding articles (I have tried to simplify
> as
> > much as possible the steps and to document them in detail).
> >
> > Best regards
> >
> > Martin de La Gorce
> > http://imagine.enpc.fr/~de-la-gm/
> >
> > --
> > IPOL - Image Processing On Line - http://ipol.im/
> >
> > contact edit at ipol.im - http://www.ipol.im/meta/contact/
> > news+feeds twitter @IPOL_journal - http://www.ipol.im/meta/feeds/
> > announces announce at list.ipol.im - http://tools.ipol.im/mm/announce/
> > discussions discuss at list.ipol.im - http://tools.ipol.im/mm/discuss/
> >
> >
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <
> https://tools.ipol.im/mailman/archive/discuss/attachments/20160505/2872b68f/attachment-0001.html
> >
>
> ------------------------------
>
> Message: 2
> Date: Thu, 05 May 2016 11:13:38 +0200
> From: Miguel Colom <colom at cmla.ens-cachan.fr>
> To: moreljeanmichel at gmail.com, general discussions about the IPOL
> project <discuss at list.ipol.im>
> Subject: Re: [IPOL discuss] A library of python bindings for IPOL
> algorithms
> Message-ID: <20160505111338.42023tfusve1wc76 at webmail.ens-cachan.fr>
> Content-Type: text/plain; charset=ISO-8859-1; DelSp="Yes";
> format="flowed"
>
> Dear all,
> yes of course, it suffices to add the link to the source code which
> includes the bindings and put it as Additional Material. It should be
> added a text explaining that it is extra material providing bindings
> and not the original peer-reviewed source code.
>
> Python is (in my very personal opinion) one of the best languages
> nowadays and it's widely used. Thus, it seems quite reasonable to
> provide such bindings.
> Also, the programs will have a Python interface to interact with them,
> and at the same time they will use fast compiled C/C++ code.
>
> Best,
> Miguel
>
> Quoting Jean-Michel Morel <moreljeanmichel at gmail.com>:
>
> > Dear Martin, and all,
> > This seems to me very important as many potential users of the algorithms
> > do not use C++.
> > Thus I would suggest to Jos? Luis and Miguel to consider including the
> link
> > to the Python interface in the web page of each paper for which it is
> ready.
> > Miguel and Jos? Luis, is that possible?
> > Best,
> > Jean-Michel
> >
> >
> > Jean-Michel Morel
> > moreljeanmichel at gmail.com
> >
> > On Wed, May 4, 2016 at 11:11 AM, martin de la gorce <
> > martin.delagorce at gmail.com> wrote:
> >
> >> Hello
> >>
> >> I have started a project on Github (
> >> https://github.com/martinResearch/PyIPOL) that aims at providing a
> python
> >> interface to each of the algorithms that are on IPOL with an example for
> >> each algorithm.
> >>
> >> I think that having Python bindings for the IPOL algorithms can be
> >> interesting to people who want to test easily and integrate quickly
> these
> >> codes in their applications written in python or in order to script
> tests
> >> easily . The goal is to put it on the official python packages index
> >> website (https://pypi.python.org/pypi) which could potentially increase
> >> the visibility of the IPOL algorithms.
> >>
> >> I have for now interfaced about a dozen of papers and the goal is to get
> >> almost all of the IPOL papers interfaced with python. Most of the
> >> interfaces have been done by calling the original executable with
> temporary
> >> files. Some bindings have been written using Cython.
> >>
> >> Adding a paper takes between 15 min and an hour, depending on the
> >> complexity of the input/output formats for the executable. Writing the
> code
> >> to get nice drawings of the results in python might take a bit of time
> when
> >> there is not yet another paper that perform a similar task already
> >> interfaced on whose code can be reused (drawing point matches for
> images of
> >> different sizes for example, or optical flow vector fields).
> >>
> >> I invite you to install a test my bindings and to make suggestions for
> >> improvements and contribute by adding articles (I have tried to
> simplify as
> >> much as possible the steps and to document them in detail).
> >>
> >> Best regards
> >>
> >> Martin de La Gorce
> >> http://imagine.enpc.fr/~de-la-gm/
> >>
> >> --
> >> IPOL - Image Processing On Line - http://ipol.im/
> >>
> >> contact edit at ipol.im - http://www.ipol.im/meta/contact/
> >> news+feeds twitter @IPOL_journal - http://www.ipol.im/meta/feeds/
> >> announces announce at list.ipol.im - http://tools.ipol.im/mm/announce/
> >> discussions discuss at list.ipol.im - http://tools.ipol.im/mm/discuss/
> >>
> >>
> >
>
>
>
>
>
> ------------------------------
>
> Message: 3
> Date: Thu, 5 May 2016 11:21:23 +0200
> From: Jos? Luis Lisani <joseluis.lisani at uib.es>
> To: moreljeanmichel at gmail.com, general discussions about the IPOL
> project <discuss at list.ipol.im>
> Subject: Re: [IPOL discuss] A library of python bindings for IPOL
> algorithms
> Message-ID: <92548aa4-37bf-1791-8e1e-8974bf1931da at uib.es>
> Content-Type: text/plain; charset="windows-1252"; Format="flowed"
>
> Hi all,
> > Thus I would suggest to Jos? Luis and Miguel to consider including the
> > link to the Python interface in the web page of each paper for which
> > it is ready.
> > Miguel and Jos? Luis, is that possible?
> Yes, sure. It can be done, we can add a link to the github project at
> https://github.com/martinResearch/PyIPOL
> But, if I understand correctly, one can not just download the wrapper
> for a single code but for the whole bundle.
> Is that right?
> In that case, maybe we could display an advertisement in the IPOL home
> page announcing the existence of the
> repository.
>
> Best,
> Jos? Luis
>
>
>
>
> El 5/5/16 a las 10:01, Jean-Michel Morel escribi?:
> > Dear Martin, and all,
> > This seems to me very important as many potential users of the
> > algorithms do not use C++.
> > Thus I would suggest to Jos? Luis and Miguel to consider including the
> > link to the Python interface in the web page of each paper for which
> > it is ready.
> > Miguel and Jos? Luis, is that possible?
> > Best,
> > Jean-Michel
> >
> >
> > Jean-Michel Morel
> > moreljeanmichel at gmail.com <mailto:moreljeanmichel at gmail.com>
> >
> > On Wed, May 4, 2016 at 11:11 AM, martin de la gorce
> > <martin.delagorce at gmail.com <mailto:martin.delagorce at gmail.com>> wrote:
> >
> > Hello
> >
> > I have started a project on Github
> > (https://github.com/martinResearch/PyIPOL) that aims at providing
> > a python interface to each of the algorithms that are on IPOL with
> > an example for each algorithm.
> >
> > I think that having Python bindings for the IPOL algorithms can be
> > interesting to people who want to test easily and integrate
> > quickly these codes in their applications written in python or in
> > order to script tests easily . The goal is to put it on the
> > official python packages index website
> > (https://pypi.python.org/pypi) which could potentially increase
> > the visibility of the IPOL algorithms.
> >
> > I have for now interfaced about a dozen of papers and the goal is
> > to get almost all of the IPOL papers interfaced with python. Most
> > of the interfaces have been done by calling the original
> > executable with temporary files. Some bindings have been written
> > using Cython.
> >
> > Adding a paper takes between 15 min and an hour, depending on the
> > complexity of the input/output formats for the executable. Writing
> > the code to get nice drawings of the results in python might take
> > a bit of time when there is not yet another paper that perform a
> > similar task already interfaced on whose code can be reused
> > (drawing point matches for images of different sizes for example,
> > or optical flow vector fields).
> >
> > I invite you to install a test my bindings and to make suggestions
> > for improvements and contribute by adding articles (I have tried
> > to simplify as much as possible the steps and to document them in
> > detail).
> >
> > Best regards
> >
> > Martin de La Gorce
> > http://imagine.enpc.fr/~de-la-gm/
> > <http://imagine.enpc.fr/%7Ede-la-gm/>
> >
> > --
> > IPOL - Image Processing On Line - http://ipol.im/
> >
> > contact edit at ipol.im <mailto:edit at ipol.im> -
> > http://www.ipol.im/meta/contact/
> > news+feeds twitter @IPOL_journal - http://www.ipol.im/meta/feeds/
> > announces announce at list.ipol.im <mailto:announce at list.ipol.im> -
> > http://tools.ipol.im/mm/announce/
> > discussions discuss at list.ipol.im <mailto:discuss at list.ipol.im> -
> > http://tools.ipol.im/mm/discuss/
> >
> >
> >
> >
> > --
> > IPOL - Image Processing On Line - http://ipol.im/
> >
> > contact edit at ipol.im - http://www.ipol.im/meta/contact/
> > news+feeds twitter @IPOL_journal - http://www.ipol.im/meta/feeds/
> > announces announce at list.ipol.im - http://tools.ipol.im/mm/announce/
> > discussions discuss at list.ipol.im - http://tools.ipol.im/mm/discuss/
>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <
> https://tools.ipol.im/mailman/archive/discuss/attachments/20160505/a8b684fa/attachment.html
> >
>
> ------------------------------
>
> Subject: Digest Footer
>
> _______________________________________________
> discuss mailing list
> discuss at list.ipol.im
> https://tools.ipol.im/mailman/listinfo/discuss
>
> ------------------------------
>
> End of discuss Digest, Vol 35, Issue 2
> **************************************
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://tools.ipol.im/mailman/archive/discuss/attachments/20160506/7a4133f4/attachment-0001.html>
More information about the discuss
mailing list