We wanted to make something that could inspire children to learn and be interested in math. Here we wanted to incorporate lego, to make it fun.
What it does
Quick mafs will create a simple mathematical question. Then the child will build the answer in lego, and place it in front of the kinect. Using tensorflow, the program will then recoqnise the number, and verify if it is the correct result.
Challenges we ran into
Kinect had a lot of noise on the depth vision, and light was a problem for the normal image. We went with the image alone, however a combination, would allow for a more open setup, as we do then not need a background. Tensorflow had issues recoqnising handwritten digits, when trained on MNIST numbers. We did not have the data to train it specifically on lego.
Accomplishments that we're proud of
Taking a picture of a 3 build in lego, parsing it to tensorflow, and having it recoqnise it as a 3
What we learned
We have learned a lot more about how to make c# and python, as well as how to have kinect and tensorflow run in those enviroments.
What's next for Quick Mafs
Getting a friendlier name Add suport for more digits Add suport for letters, so it can be used to learn spelling as well