Print

[zetathix]

Now I'm still reading documents from Innvo and Compuphase. What's interesting is a sensor that uses to communicate with head-cpu (I will call this HC) and main-CPU (MC) and capable to grab raw data from camera. But anyone ever try it yet?

[irbyma7297]

I've heard about taking pictures with pleo s camera. Is that what you are talking about?

[zetathix]

Umm.. Not really quite a "native bool: camera_take_picture(const file_name[]);".

Something like "SENSOR_PACKET        =    44, // virtual sensor to get the packet data from the NXP processor"

That maybe uses with ...     "PACKET_CAMERA  = 4,      // 1 <<  2, // 'c'", like on the PPG's example.

(I don't know this's still using in rb or not, but last time I checked on "include\pleo\sensors.inc", it's still there.)


But on "include\Sensor.inc". For camera, it already has...

native sensor_data_size(sensor_name: sensor); // to know what data size that need to contain.

native sensor_read_data(sensor_name: sensor, data[], length = sizeof data);  //  to read raw data on the camera which return in array cell/s. This's a comment here "*  @todo add a packed parameter", I guess that's still doesn't have.


my only purpose now is just try to figure out how Pleo could do SURF or SIFT method, with itself alone?, or need to throw the data out and crunching with some external CPU like smartphone. Still don't have a clue.
But this is only a thinking process and I'm still didn't capable to adopt a Pleo. I just think it doesn't need a penny to keep thinking...

Note: my idea maybe too naive for hardware that's only 64KB for fast speed 1 cycle.

[kat]

There aren't very many techy forum members among us unfortunately so most of us don't really have a clue what your talking about.  There are a few around though so maybe wait for them to pop in and see what they think.    

[mlbelanger]

I've just started investigating the camera API but I haven't got around to the generic SENSOR_PACKET sensor yet.  In theory you'd be able to read raw data out of the camera or the microphones instead of using the built-in API routines but I have a feeling the limiting factor will be the slow speed of the serial connection between the head and body CPUs.

You can also have the camera report back an average value for a region of the screen so if you wanted to for instance get a simple 3x3 pixel image from the camera breaking up the screen into 9 quadrants that would be doable but I'm sure it would take forever to deliver the data.

I think to truly do custom image recognition routines on the Pleo we'll have to wait until more information becomes available in future PDK releases.  Perhaps some day we'll be able to custom compile machine language routines directly for the head CPU but in the mean time we're very limited with what we can do.

[zetathix]

Please don't worry that this is too technical, I'm still confuse and all I say maybe totally wrong and when I read what I posted, that's a mess. 

For now, sometime I think that Innvo hold the developement to let Pleo prove its value in the marget. There's no new FW come out for the long 1 or 2 years. It feels like pausing. 

I have try to asking for the news about bench-adapter and more-polished PPG and API on Pleo-fb page. I will let you all know if I receive any response.