Baseline Challenge

majority guess vs. real k-NN vote Not connected

Hardware

Sensor inputs

Tick every reading the model should compare on:

Simulation inputs

Motor position (training)

Mode TRAINING

Train a labeled example

Type a label name (e.g. "Cat" or "Bin A"), hand-position the arm/sensor for that example, then record it.

Live

Sensor: —
Motor: —
Recorded examples: 0

Examples per label (imbalance shows up here)

The Real Model looks at the 3 closest examples and votes. The Baseline ignores the sensor entirely and always guesses whichever label has the most examples.

Run mode — compare the guesses

Baseline guess

0 / 0

Real model guess

0 / 0
The arm physically moves to the real model's guess. The baseline is shown for comparison only. Tap Correct/Wrong after each new guess to track accuracy — try training a lopsided dataset (lots of one label, barely any of another) and watch the baseline's accuracy hold up misleadingly well, until you test the minority label.