Classifier settings

k (neighbours voted): 5 Damage vote threshold: 0.50 Scan speed: 6.0 patches/s

Leaf sample

Current patch

Patch—

Mean intensity—

Local variance—

Edge density—

Vote—

Confidence—

Scan stats

Patches classified healthy0

Patches classified damaged0

Pass progress0 / 0

Each patch's real pixel data is reduced to three genuine texture features — mean intensity, local variance and Sobel edge-density — then classified by a k-NN vote (real Euclidean distance) against a labelled reference set of healthy and pest-damaged patches sampled from this same leaf.