Abstract:
We present LUCY-learn, an application designed for use in classrooms, that allows students to experi- ment with the application of different image classifiers to satellite images in the vicinity of the students’ school or neighbourhood. The Land Use Classification sYstem (LUCY) illustrates fundamental proba- bilistic principles of machine learning by visualising the probability distributions of classifier features and output layers. Learning is scaffolded through visualisations ranging from a hue-based pixel classi- fier to comparison of local image features acquired by a simple convolutional neural network. The result has been evaluated in classroom trials, with successful learning outcomes.
PPIG 2025 - 36th Annual Workshop
LUCY-learn: Visualising Image Classifiers for Education in Machine Learning