After startup weekend and looking at shopsavvy, I started thinking about image recognition. shopsavvy uses a cell phone’s camera to read barcodes. cool. There is a lot that could be done with that. But how about something more useful?
I’d like to take a netbook (like the Asus EEE) and a web-cam (or maybe more than 1) and see what could be done. For instance: put it in a car. Have 2 cameras, one facing forward and one facing backward. What could be done? Can we grab license plates, recognize the numbers and then get information? There are already products for this that the police use. But for private citizens it might not be bad. For instance a pull from the DMV might tell you the driver has a lot of accidents and DWIs. Perhaps you should give them some space.
Or privacy: how many cameras are pointed at you right now? Wear the rig with the cameras facing forward and backward. As you walk it tells you where the cameras are. Probably great for bank robbers. But criminals aren’t the only people that value their privacy.
You also have the simple application of knowing who you are talking to. If you can recognize a person’s face and know whether or not you’ve encountered this person before would be nice. Perhaps it’s PRM instead of CRM. But no more stammering around as you search your memory.
Right now I’m not concerned about the display device. Just being able to have a single application be able to recognize stuff in real time should be challenging enough. I won’t even touch on speech and sound recognition.