= Colour Calibration =
== Set Player and install to memory stick ==
In {{{robot}}} directory run:
{{{
./spip 1 1 ip
}}}
the first "1" is jersey number 1, the second "1" is c++ player 1 - the photo taker player. see [wiki:spip spip] for more details.
now run:
{{{
./ins
}}}
to install the files to the memory stick
== Taking photos ==
insert the memory stick into the dog and wait until it boots. you can now take photos by pressing the head button (it takes a photo when the LEDs change). the photos are stored to the root directory of the memory stick. eg:
YUV00.BFL?BR
YUV01.BFL?BR
...
== Painting images ==
You train the aibo to recognize colours by painting regions of the image with colour. Create a directory under {{{base/colour/}}} for your images. For example, from your trunk directory:
{{{
mkdir base7/colour/20040902
}}}
in {{{base/colour/ic}}} run:
{{{
java ic
}}}
This opens up a tool that you can paint images with. (add more on this later ...tom)
== Creating a calibration file for the rUNSWift AIBO ==
after classifying and saving all images, run in {{{base/colour/classify}}}:
{{{
./dtcal.sh ../20040902
}}}
or more generally
{{{
./dtcal.sh [directory where the painted bfl files are]
}}}
(NOTE: dtcal.sh is a fixed version of dtcalibrate.sh which was brittle. it can use any relative directory name)
This should not produce any errors, and will create a colour calibration file at {{{files/nnmc.cal}}}. You need to copy this into your robot7/cfg directory and then upload to the memory stick. for example:
{{{
cp files/nnmc.cal <trunk root>/robot/cfg/
cd <trunk root>/robot/
./spip 1 Forward 141 (if you want to test out the forward with the new colour calibration)
./ins
}}}