Tiny Cats Model — Classifier Demo

Binary cat vs dog classifier (ResNet-18, 1478 images per split eval, Oxford IIIT Pet), trained with class-balanced loss on the full dataset. Evaluation report in evaluation_report.json at the repo or the model hub repo.

Headline metrics

Metric	Value
Accuracy	99.66% (1473/1478)
Macro F1	99.61%
Weighted F1	99.66%
Misclassifications	5

Per-class precision / recall / F1

Class	Precision	Recall	F1
cat	99.58%	99.37%	99.48%
other	99.70%	99.80%	99.75%

Confusion matrix (rows=true, cols=predicted)

	cat	other
cat	476	3
other	2	997

Report timestamp: 2026-08-12T10:51:31.480493+00:00