Deploy FaceGuard demo with proper README config

README.md

@@ -1,3 +1,14 @@
+---
+title: FaceGuard – ViT Demo
+emoji: 🎭
+colorFrom: indigo
+colorTo: pink
+sdk: gradio
+sdk_version: "4.0.0"
+app_file: app.py
+pinned: false
+---
+
 # FaceGuard – ViT (20 CelebA IDs)
 
 A lightweight face-identity classifier demo.
@@ -13,10 +24,8 @@ This Space lets you upload a face crop and see **Top-5 predicted identities**.
 *Note:* CelebA identities are anonymous integer IDs (e.g., 8968), not real names.
 
 ## Visuals
-If training artifacts were produced in Colab, they’re included below:
 ![Confusion Matrix](./confusion_matrix_test.png)
 ![ROC Curves](./roc_curves_test.png)
 
 ## Tech
 - 🤗 Transformers (ViT), 🤗 Datasets, Gradio UI
-- Collator-based preprocessing (no `with_transform`) for robust vision batching

app.py

@@ -21,7 +21,6 @@ def predict_top5(img: Image.Image):
         logits = model(pixel_values=enc["pixel_values"]).logits
         probs = torch.softmax(logits, dim=-1).squeeze(0).cpu().numpy()
     top5 = probs.argsort()[-5:][::-1]  # indices of Top-5 classes
-    # Gradio Label expects a dict of {display_name: score}
     return {f"label {i} (celeb_id {id2label[i]})": float(probs[i]) for i in top5}
 
 desc = (