Update README.md

README.md

@@ -18,6 +18,9 @@ Try 3 backbones (ViT / ResNet / EfficientNet) on any image. We show **top-5 pred
 - Pick a model in the dropdown
 - Click **Predict**
 
+## Why this matters
+Different backbones behave differently. This demo makes those trade-offs visible to judges/users in one place.
+
 ## Roadmap
 - Eval notebook → Confusion Matrix + Accuracy + ROC (binary subsets)
 - Auto-publish plots to this README