Quantcast

[Scilab-users] Scilab Computer Vision Module

classic Classic list List threaded Threaded
6 messages Options
Tan Chin Luh-2 Tan Chin Luh-2
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

[Scilab-users] Scilab Computer Vision Module

Hi,

It is great to have the new computer vision released on atoms, especially a module which is created by Scilab Enterprises.

After so long there are many image processing and computer vision modules in atoms, which have their pros and cons, and this new module finally allows the call of almost all functions (if not all :) ) to the OpenCV 2.4.13 through its gateway function although not all are documented. I believe it will be coming very soon.

To make it better I think we could share some of our finding of the "hidden" functions" while waiting for the module update, the following code perform the image inpaint algo:

scicv_Init();

S = imread(getSampleImage("blobs.jpg"));

// Creating mask to remove the object at 2nd row, 2nd col. This would be easier after the support of the insertion for Mat implemented as stated in help file
m1 = uint8(zeros(Mat_rows_get(S),Mat_cols_get(S)));
m1(110:200,150:250) = 255;
imwrite('mymask.jpg',m1);

mask = imread('mymask.jpg',CV_LOAD_IMAGE_GRAYSCALE);

S2 = inpaint(S,mask,3,INPAINT_NS);
matplot(S2)



Regards,
Chin Luh








_______________________________________________
users mailing list
[hidden email]
http://lists.scilab.org/mailman/listinfo/users
Simon Marchetto Simon Marchetto
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Scilab Computer Vision Module

Hello,

Thanks very much for your support.

I was also very impressed about the new IPCV module, the number and
quality of available functions ! Particularly those related to GUI
manipulation, that we really missed, in comparison with Matlab.
We really appreciate your work.

For years we have been wanting to provide and support an image
processing toolbox, because this kind of toolboxes is among the most
used, at least downloaded, by the Scilab community.

Yes indeed there are some image processing toolboxes now, especially
based on OpenCV, but no so much, if you compare with the huge number
available for Python.
For me several (good) toolboxes is definitely better than one at this
moment. And they can have all different purposes, and follow different
choices in design.

About hidden functions in sciCV, they will be more and more documented,
version after version.

Regards,

Simon Marchetto


Le 2017-04-25 05:52, Tan Chin Luh a écrit :

> Hi,
>
> It is great to have the new computer vision released on atoms,
> especially a module which is created by Scilab Enterprises.
>
> After so long there are many image processing and computer vision
> modules in atoms, which have their pros and cons, and this new module
> finally allows the call of almost all functions (if not all :) ) to
> the OpenCV 2.4.13 through its gateway function although not all are
> documented. I believe it will be coming very soon.
>  To make it better I think we could share some of our finding of the
> "hidden" functions" while waiting for the module update, the following
> code perform the image inpaint algo:
>
> scicv_Init();
>
> S = imread(getSampleImage("blobs.jpg"));
>
> // Creating mask to remove the object at 2nd row, 2nd col. This would
> be easier after the support of the insertion for Mat implemented as
> stated in help file
> m1 = uint8(zeros(Mat_rows_get(S),Mat_cols_get(S)));
> m1(110:200,150:250) = 255;
> imwrite('mymask.jpg',m1);
>
> mask = imread('mymask.jpg',CV_LOAD_IMAGE_GRAYSCALE);
>
> S2 = inpaint(S,mask,3,INPAINT_NS);
> matplot(S2)
>
>  Regards,
>  Chin Luh
>
>
> _______________________________________________
> users mailing list
> [hidden email]
> http://lists.scilab.org/mailman/listinfo/users
_______________________________________________
users mailing list
[hidden email]
http://lists.scilab.org/mailman/listinfo/users
Tan Chin Luh-2 Tan Chin Luh-2
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Scilab Computer Vision Module

Hi Simon,

Thanks for your comments. Just to add, IPCV is modified from SIVP, with
extra features. 1-2 functions were taken from IPD and SIP quite some
time ago.

 From my opinion, the common issues of all the module above are :
1. Distributing the binary cross platform, which always give the issue
of lib version.
2. Functions from the OPENCV are "link" to Scilab with individual
gateway, which lead to the limited functions depending on the developer.
3. Data passing and sharing between Scilab and OpenCV

 From the look of it, the scicv seems to have solved the first 2 issues,
and near to 1k functions in OPENCV seems to have included in the dll,
which I believe it break the bottle neck of which others (including me)
facing. I wonder was it hand-coded? As for number 3, I believe everyone
might have different opinion.

Yes you're right, for the moment there should be more modules, as OpenCV
functions are being used in different functions to perform a high level
task in Scilab, which I believe all developers would have their own way
to do so. Eventually I believe once the framework/basic have setup, we
could move towards it.

Just my 2 cents, sorry if any of my statements sound wrong.

Thanks.

Regards,
Chin Luh



On 25/4/2017 4:35 PM, [hidden email] wrote:

> Hello,
>
> Thanks very much for your support.
>
> I was also very impressed about the new IPCV module, the number and
> quality of available functions ! Particularly those related to GUI
> manipulation, that we really missed, in comparison with Matlab.
> We really appreciate your work.
>
> For years we have been wanting to provide and support an image
> processing toolbox, because this kind of toolboxes is among the most
> used, at least downloaded, by the Scilab community.
>
> Yes indeed there are some image processing toolboxes now, especially
> based on OpenCV, but no so much, if you compare with the huge number
> available for Python.
> For me several (good) toolboxes is definitely better than one at this
> moment. And they can have all different purposes, and follow different
> choices in design.
>
> About hidden functions in sciCV, they will be more and more
> documented, version after version.
>
> Regards,
>
> Simon Marchetto
>
>
> Le 2017-04-25 05:52, Tan Chin Luh a écrit :
>> Hi,
>>
>> It is great to have the new computer vision released on atoms,
>> especially a module which is created by Scilab Enterprises.
>>
>> After so long there are many image processing and computer vision
>> modules in atoms, which have their pros and cons, and this new module
>> finally allows the call of almost all functions (if not all :) ) to
>> the OpenCV 2.4.13 through its gateway function although not all are
>> documented. I believe it will be coming very soon.
>>  To make it better I think we could share some of our finding of the
>> "hidden" functions" while waiting for the module update, the following
>> code perform the image inpaint algo:
>>
>> scicv_Init();
>>
>> S = imread(getSampleImage("blobs.jpg"));
>>
>> // Creating mask to remove the object at 2nd row, 2nd col. This would
>> be easier after the support of the insertion for Mat implemented as
>> stated in help file
>> m1 = uint8(zeros(Mat_rows_get(S),Mat_cols_get(S)));
>> m1(110:200,150:250) = 255;
>> imwrite('mymask.jpg',m1);
>>
>> mask = imread('mymask.jpg',CV_LOAD_IMAGE_GRAYSCALE);
>>
>> S2 = inpaint(S,mask,3,INPAINT_NS);
>> matplot(S2)
>>
>>  Regards,
>>  Chin Luh
>>
>>
>> _______________________________________________
>> users mailing list
>> [hidden email]
>> http://lists.scilab.org/mailman/listinfo/users
>

_______________________________________________
users mailing list
[hidden email]
http://lists.scilab.org/mailman/listinfo/users
Antoine Monmayrant Antoine Monmayrant
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: ?==?utf-8?q? Scilab Computer Vision Module

>  From the look of it, the scicv seems to have solved the first 2 issues,
> and near to 1k functions in OPENCV seems to have included in the dll,
> which I believe it break the bottle neck of which others (including me)
> facing. I wonder was it hand-coded?

They might have used SWIG ( http://www.swig.org/ ) or some similar tool to automate the generation of the binding code.

Antoine

_______________________________________________
users mailing list
[hidden email]
http://lists.scilab.org/mailman/listinfo/users
Tan Chin Luh-2 Tan Chin Luh-2
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: ?==?utf-8?q? Scilab Computer Vision Module

Hi,

I am not familiar with SWIG, so I though the binding will still need
some hand coding as well. thanks for the link.

regards,
Chin Luh


On 26/4/2017 3:00 PM, Antoine Monmayrant wrote:

>>   From the look of it, the scicv seems to have solved the first 2 issues,
>> and near to 1k functions in OPENCV seems to have included in the dll,
>> which I believe it break the bottle neck of which others (including me)
>> facing. I wonder was it hand-coded?
> They might have used SWIG ( http://www.swig.org/ ) or some similar tool to automate the generation of the binding code.
>
> Antoine
>
> _______________________________________________
> users mailing list
> [hidden email]
> http://lists.scilab.org/mailman/listinfo/users
>

_______________________________________________
users mailing list
[hidden email]
http://lists.scilab.org/mailman/listinfo/users
Simon Marchetto Simon Marchetto
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: ?==?utf-8?q? Scilab Computer Vision Module

Hello,

Yes indeed we use SWIG.

As you said, it needs hand coding.

First there is a couple of issues in the OpenCV code for SWIG, it needs
some code patching.
Then it needs some "cleaning" to get the desired code wrapped.

Maybe because of these two issues, OpenCV abandoned SWIG for their
Python bindings, and used a custom wrapping method.
We couldn't afford to do that.

After this, you need to implement to typemaps (SWIG code to convert the
OpenCV data types from/to Scilab).

All of the work is not finished yet.

Simon

Le 2017-04-26 10:55, Tan Chin Luh a écrit :

> Hi,
>
> I am not familiar with SWIG, so I though the binding will still need
> some hand coding as well. thanks for the link.
>
> regards,
> Chin Luh
>
>
> On 26/4/2017 3:00 PM, Antoine Monmayrant wrote:
>>>   From the look of it, the scicv seems to have solved the first 2
>>> issues,
>>> and near to 1k functions in OPENCV seems to have included in the dll,
>>> which I believe it break the bottle neck of which others (including
>>> me)
>>> facing. I wonder was it hand-coded?
>> They might have used SWIG ( http://www.swig.org/ ) or some similar
>> tool to automate the generation of the binding code.
>>
>> Antoine
>>
>> _______________________________________________
>> users mailing list
>> [hidden email]
>> http://lists.scilab.org/mailman/listinfo/users
>>
>
> _______________________________________________
> users mailing list
> [hidden email]
> http://lists.scilab.org/mailman/listinfo/users
_______________________________________________
users mailing list
[hidden email]
http://lists.scilab.org/mailman/listinfo/users
Loading...