The Zumi Robot is a small, programmable self-driving car kit that uses AI to learn and grow. The goal of using the Zumi Robot in this course is to train it in obstacle detection and avoidance.
Building Process: Building the Zumi robot was a pretty smooth process. We followed the instructions and there were not many challenges besides the battery running out quickly.
Training the Zumi to detect colors was a pretty simple process. We first labeled which color we were about to assign and then would hold up that color. As the color card was held up to the screen thirty photos were taken. This ensured that different lighting or distances affected the robots ability to detect that color. After completing this process for each color we tested Zumi. We held up each color and Zumi was able to read it just as we programmed.
To get the Zumi to be able to translate hand signals into motion we had to train it. First, one person created three hand signals representing forward, right, and left. That person held up a signal while the computer took at least twenty pictures to ensure accuracy. For example, to represent the signal "forward" we chose to face one palm to the camera. Then we were ready to test. Zumi had trouble knowing the difference between signals "right" and "left." To fix this issue, we redid the hand signals to differentiate them more. This worked and we successfully were able to navigate our Zumi.
Results: We successfully built the Zumi. She is able to detect colors and be navigated through hand gestures.