Pi-ggy Bank
A money box that counts your coins with computer vision. 1st place, Best Hack at hackSheffield 5.
Pi-ggy Bank is a money box that uses computer vision to work out which coins are dropped into it, so you can keep track of how much you have saved. We built Pi-ggy Bank at hackSheffield 5, where it won Best Hack (1st place). It was our first hardware hack.
How it works
- Find the coins: OpenCV segments each image to locate the coins.
- Describe them: each coin is cut out of the image and its histogram features are computed.
- Train: a few hundred training images, extracted with the first stage, are used to train a Support Vector Machine (SVM) classifier with scikit-learn.
- Predict: when a coin is dropped, a series of images is captured, the classifier predicts the coin in each one, and the most common prediction wins.
Lighting was the main challenge: training and test images were taken at different times of day, so the lighting conditions are normalised before classification.
My contribution
I collected the training data, built the physical prototype, and used OpenCV to compute the histogram features of the coins.