KI-Training
AI training promotes a basic understanding of data, training, sources of error, and decision-making processes in AI systems. With simple exercises and projects in the field of machine learning, movement patterns are recorded, your own models are trained, and used in prototypes. The Calliope mini enables visible control and an engaging way to process data and make it tangibly experienceable. calliope
How machines learn
How does a model’s behavior change when you add new or different data?
How do faulty, incomplete, or biased data influence the result of an AI model?
How does the reliability of a prediction change when you collect more training data?
How it works
Prepare the Calliope mini
The Calliope mini is connected to the computer via Web-USB. The ML program is then downloaded to the Calliope mini via the website. docs.calliope
⚠️ KI-Training requires a browser that supports Web-USB, e.g. Chrome or Edge. makecode.calliope
Connect the Calliope mini
The Calliope mini is connected to KI-Training via Web-Bluetooth. The connection is checked by comparing the pattern shown on the LED matrix with the pattern displayed in the browser. docs.calliope
Generate data
Using the Calliope mini, different movement patterns (e.g. tilting, shaking) are recorded and saved as data sets. calliope
Train the model
Based on the collected data, a machine learning model is created that can distinguish between the movement patterns. docs.calliope
Test and review the model
The algorithm’s decisions are displayed directly so they can be checked and reflected on. docs.calliope
Apply the model
The trained model is integrated into a MakeCode program and transferred to the Calliope mini. The algorithm then runs autonomously on the device and controls its own application. docs.calliope
