Update README - Run 20251012_041353

README.md

@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 82.08
+      value: 82.20
 ---
 
 # David: Multi-Scale Crystal Classifier
@@ -51,15 +51,15 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 82.08%
-- **Best Epoch**: 4
-- **Final Train Accuracy**: 86.83%
+- **Validation Accuracy**: 82.20%
+- **Best Epoch**: 5
+- **Final Train Accuracy**: 87.96%
 
 ### Per-Scale Performance
-- **Scale 384**: 82.08%
-- **Scale 768**: 82.11%
-- **Scale 1024**: 82.25%
-- **Scale 1280**: 81.95%
+- **Scale 384**: 82.20%
+- **Scale 768**: 82.33%
+- **Scale 1024**: 82.33%
+- **Scale 1280**: 82.18%
 
 
 ## Usage
@@ -182,4 +182,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-10-12 04:25:22*
+*Generated on 2025-10-12 04:27:42*