I am also trying to figure out the AVRcam. I upload the color map using the software from the CD, AVRcamView. I take a picture of the objects, then point with the mouse at the color I want to add to the colour map, right click and select add to map, then click on one of the 8 locations and click on send. Then go back to the picture, select another color, add it to a different location, send again.
Now I have a question for you, something that I didn't figure out yet. How do you parse the bounding boxes? I mean how do you find out how to turn the robot and have it go towards the object you want?