The problem Ball Following Robo-Car solves

This bot uses pure computer vision concepts. Compared to conventional ML models, it:
1.Has a higher FPS.
2.Has a lower computational complexity. (Works well on edge devices like the Raspberry Pi)
3.Need not be trained.

Future Applications :
The project can be extended to sense human heat using heat sensor.
Machine learning models can be implemented to better train the robot to track unusual or out of the place objects or people in a busy environment.
Image analysis part can be used for home automated security systems, automated CCTV’s which can track intruders and click pictures and sent it over wireless system.

Challenges I ran into

The biggest problem was the inacuracy of the motors whose power supply changed after every few uses thus troubling us with changing the speed of motors through code. Moreover, each motor was changing differently hence each motor had to be calliberated differently. Grease also had to be applied to the servo motor.
Calliberating the ball differently according to it's colour and the light anvironment everytime was a cumbersome task as it is based on shape and colour detection.