Take away fear of computers/robots and help humans understand how computers think.
Let's two groups of people compete against each other and learn how computers understand the world by playing with them.
In order to do that, a team chooses one object present in a scene which both teams need to describe to Roboy. Each team have to describe the object without using the word itself and user defined "stop words".
If a team describes it accurate enough for Roboy to get what the object was, it'll get one point. Using prohibited words results in losing a point.
Using the Roboy DialogSystem in Java we implemented state machine. Speech recognition was built with Google Speech Recognition. Text-to-Speech was implemented with Google Text-to-Speech (gTTS 1.1.6).
Image captured by Kinect sensor is then analysed using Darknet for ImageNet in order to recognize objects present in the frame. Word Associations helped us letting the computer understand the meaning of the spoken word.
All components communicate using ROS. Intel Edison is used to show in detail how good the object was described.
- Challenges we ran into: Memory limitations of Intel-Edison. Integrating all the components to work with each other.
- Accomplishments that we're proud of: Making it all work because it is a new idea and a lot of fun.
- What we learned: Programming and working with Edison.
- What's next for TheGreatGuessingGame: Porting it to work on Roboy. Finalizing emotions. Implementing movement.