"You are stuck in an unknown place. No GPS, only a connection to your friend who knows where to go."
Using Azure Cloud Services, the user can submit directions using natural speech, such as "Keep on going ahead" or "You should go back" or "Go to the left now". We trained a model to detect intent from the voice commands, and return simplified commands to the local server. The server then renders the maze with the new location of the trapped friend. The server also does 2D collision detection to ensure that the friend does not "walk through walls" like some ghost :P.
This project can be scaled by pulling from maze generators of indeterminate size, and can be adapted to gain intent from more natural speech with a longer trained voice model.
Log in or sign up for Devpost to join the conversation.