Andy

An incredibly cheeky question but hey, if you don't ask you don't get and I'm running out of time on a project.

I know this has been discussed many times by various people but.... does anybody have a simple working shape recognition system made with cv.jit objects that they would be willing to let me take a look at or describe the processes for - or any good papers suitable for someone with max knowledge but not necessarily any C knowledge?

I have looked at a lot of papers but I'm no closer to getting a working system in max. I have tried things like cv.jit.learn etc but I cannot seem to get it to demonstrate any consistency. I'm assuming there must be some levels of pre-processing that I must do before calculating the moments?

(Basically I want to recognise simple shapes drawn into an LCD - not find shapes in a complex image)

I know this has been discussed many times by various people but.... does anybody have a simple working shape recognition system made with cv.jit objects that they would be willing to let me take a look at or describe the processes for - or any good papers suitable for someone with max knowledge but not necessarily any C knowledge?  

i-know-this-has-been-asked-a-thousand-times-but

i don't think the question is cheeky, and i don't think the answer is trivial. shape recognition is a very complicated affair.

i'm not sure if you're asking for some pattern recognition algorithms, or how to make your video input better for the cv objects. if it's the former, i don't know how much advice you will get. if it's the latter, you should get plenty of advice by searching the forum.

you also might want to check out the reactable application.

i don't think the question is cheeky, and i don't think the answer is trivial.  shape recognition is a very complicated affair.

i'm not sure if you're asking for some pattern recognition algorithms, or how to make your video input better for the cv objects.  if it's the former, i don't know how much advice you will get.  if it's the latter, you should get plenty of advice by searching the forum.

you also might want to check out the reactable application.


As it happens my project includes reactivision so I'm aware of the principles behind how that particular system works.

The system I want would be very different - I want to be able to draw a free hand square into an LCD and have my system be able to distinguish between a square and a triangle for example - not find shapes in an image.

More research has told me that perhaps what I need to research is more aligned to sketch recognition rather than image analysis and I have since found an interesting paper that I am looking into - In case anyone else is interested in this subject:

http://rationale.csail.mit.edu/publications/Sezgin2001Sketchbased.pdf

If I get something working I'll post a patch up - I'm sure there are others out there trying to do the same sort of thing.

In the mean time if anyone has any experience in this area and wishes to share knowledge it would be greatly appreciated.

http://rationale.csail.mit.edu/publications/Sezgin2001Sketchbased.pdf 

You should checkout a FTM/mnm library, there's an example that does 

what you describe - simple gesture recognition. It should be 

somewhere in the Extras/ftm/mnm something menu.

> The system I want would be very different - I want to be able to 

> draw a free hand square into an LCD and have my system be able to 

> distinguish between a square and a triangle for example - not find 

You should checkout a FTM/mnm library, there's an example that does  
what you describe - simple gesture recognition. It should be  
somewhere in the Extras/ftm/mnm something menu.

> The system I want would be very different - I want to be able to  
> draw a free hand square into an LCD and have my system be able to  
> distinguish between a square and a triangle for example - not find  
> shapes in an image.


It's in the examples folder of the unzipped FTM files (from inside 

I am not having much luck getting it to work properly, though. It 

looks really interesting, are there any more examples for this out 

there? It mentions using the Wii remote, for instance.

On Nov 17, 2007, at 12:40 PM, nesa wrote:

> You should checkout a FTM/mnm library, there's an example that does 

> what you describe - simple gesture recognition. It should be 

> somewhere in the Extras/ftm/mnm something menu.

>> The system I want would be very different - I want to be able to 

>> draw a free hand square into an LCD and have my system be able to 

>> distinguish between a square and a triangle for example - not find 

It's in the examples folder of the unzipped FTM files (from inside  
the disk image).
I am not having much luck getting it to work properly, though.  It  
looks really interesting, are there any more examples for this out  
there?  It mentions using the Wii remote, for instance.

> Hi,
>
> You should checkout a FTM/mnm library, there's an example that does  
> what you describe - simple gesture recognition. It should be  
> somewhere in the Extras/ftm/mnm something menu.
>
> Cheers,
> nesa
>
>> The system I want would be very different - I want to be able to  
>> draw a free hand square into an LCD and have my system be able to  
>> distinguish between a square and a triangle for example - not find  
>> shapes in an image.
>


This sounds really interesting but I'm not sure where to find the files?

Well, I think that ftm 1.x example(ez-hmm-something) didn't work for 

me until I changed who knows what, but now I have 2.x installed, and 

mnm.follower-example works out of the box(after reading p help 

There's also FTM mailing list and a wiki page, I think thats where 

On Nov 17, 2007, at 6:58 PM, evan.raskob [lists] wrote:

> It's in the examples folder of the unzipped FTM files (from inside 

> I am not having much luck getting it to work properly, though. It 

> looks really interesting, are there any more examples for this out 

> there? It mentions using the Wii remote, for instance.

> On Nov 17, 2007, at 12:40 PM, nesa wrote:

>> You should checkout a FTM/mnm library, there's an example that 

>> does what you describe - simple gesture recognition. It should be 

>> somewhere in the Extras/ftm/mnm something menu.

>>> The system I want would be very different - I want to be able to 

>>> draw a free hand square into an LCD and have my system be able to 

>>> distinguish between a square and a triangle for example - not 

Well, I think that ftm 1.x example(ez-hmm-something) didn't work for  
me until I changed who knows what, but now I have 2.x installed, and  
mnm.follower-example works out of the box(after reading p help  
subpatch)...

There's also FTM mailing list and a wiki page, I think thats where  
Wii remote example is...

> It's in the examples folder of the unzipped FTM files (from inside  
> the disk image).
> I am not having much luck getting it to work properly, though.  It  
> looks really interesting, are there any more examples for this out  
> there?  It mentions using the Wii remote, for instance.
>
> Best
> Evan
>
>
>
> On Nov 17, 2007, at 12:40 PM, nesa wrote:
>
>> Hi,
>>
>> You should checkout a FTM/mnm library, there's an example that  
>> does what you describe - simple gesture recognition. It should be  
>> somewhere in the Extras/ftm/mnm something menu.
>>
>> Cheers,
>> nesa
>>
>>> The system I want would be very different - I want to be able to  
>>> draw a free hand square into an LCD and have my system be able to  
>>> distinguish between a square and a triangle for example - not  
>>> find shapes in an image.
>>


It does "work" but not very well. I think I need those magic numbers.

I found a few patches with FTM wiiremote examples, but nothing using 

gesture analysis. Any FTM people out there on the jitter list?

On Nov 17, 2007, at 10:37 PM, nesa wrote:

> Well, I think that ftm 1.x example(ez-hmm-something) didn't work 

> for me until I changed who knows what, but now I have 2.x 

> installed, and mnm.follower-example works out of the box(after 

> There's also FTM mailing list and a wiki page, I think thats where 

> On Nov 17, 2007, at 6:58 PM, evan.raskob [lists] wrote:

>> It's in the examples folder of the unzipped FTM files (from inside 

>> I am not having much luck getting it to work properly, though. It 

>> looks really interesting, are there any more examples for this out 

>> there? It mentions using the Wii remote, for instance.

>> On Nov 17, 2007, at 12:40 PM, nesa wrote:

>>> You should checkout a FTM/mnm library, there's an example that 

>>> does what you describe - simple gesture recognition. It should be 

>>> somewhere in the Extras/ftm/mnm something menu.

>>>> The system I want would be very different - I want to be able to 

>>>> draw a free hand square into an LCD and have my system be able 

>>>> to distinguish between a square and a triangle for example - not 

It does "work" but not very well.  I think I need those magic numbers.

I found a few patches with FTM wiiremote examples, but nothing using  
gesture analysis.  Any FTM people out there on the jitter list?

> Well, I think that ftm 1.x example(ez-hmm-something) didn't work  
> for me until I changed who knows what, but now I have 2.x  
> installed, and mnm.follower-example works out of the box(after  
> reading p help subpatch)...
>
> There's also FTM mailing list and a wiki page, I think thats where  
> Wii remote example is...
>
>
> On Nov 17, 2007, at 6:58 PM, evan.raskob [lists] wrote:
>
>> It's in the examples folder of the unzipped FTM files (from inside  
>> the disk image).
>> I am not having much luck getting it to work properly, though.  It  
>> looks really interesting, are there any more examples for this out  
>> there?  It mentions using the Wii remote, for instance.
>>
>> Best
>> Evan
>>
>>
>>
>> On Nov 17, 2007, at 12:40 PM, nesa wrote:
>>
>>> Hi,
>>>
>>> You should checkout a FTM/mnm library, there's an example that  
>>> does what you describe - simple gesture recognition. It should be  
>>> somewhere in the Extras/ftm/mnm something menu.
>>>
>>> Cheers,
>>> nesa
>>>
>>>> The system I want would be very different - I want to be able to  
>>>> draw a free hand square into an LCD and have my system be able  
>>>> to distinguish between a square and a triangle for example - not  
>>>> find shapes in an image.
>>>
>


It seems to work ok for me. It distinguishes quite well between simple shapes so long as they are drawn in the same direction as the model. Also, lots (20+) of models seem to cause 'bad' slow down. Still.....seems pretty good. Not 100% perfect, but good none the less.

I know this has been asked a thousand times but....