Only needs tensorflow for color and posture recognition, why would you need facial recognition?! I get it but it’s so much work. Just do color and posture recognition to check objects color then posture and shoot?
Mount it on a top hat or umbrella and program it to shine camera lenses to protect privacy and market it to celebs. Laser domes on car hoods/roofing would be good too. Today, we there is no privacy the moment you step out the door; the person who solves that will be bay-zoes-rich.
ChatGPT is mainly used for mistakes in code though, because it it codes something you do not understand, it does not count as “your work”, and in the future it might not be there, your choice though.
How can I provide a command to the pant tilt dual-axis servos to position them at the appropriate angles for the laser to accurately target an intended object?
That’s a difficult problem with some work arounds. You could program some kinematic vector equations (ex. What angle does the laser need to point at from an origin in order for the vector to touch a specific point on another chord system), but this is difficult because angled surfaces in the room would each require their own 3D equations. Instead, use PID control equations. Ex, I want the laser to be “here” but the laser is “there” it is off from its target by 50 px in the y and -40 px in the left. The Pid control equation would basically keep working until both of those number are zero. This is the easier solutions but takes more tuning (changing the constants in your transfer function), but the experimentation would be fun!
@@PK-wo4vo Have you considered a "calibration" phase, to identify the unknown of the coordinate mapping? That would be an alternative to the PID that requires the laser dot to be visibile at all times.
