Update README - Run 20251012_050214

README.md

@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 81.82
+      value: 82.10
 ---
 
 # David: Multi-Scale Crystal Classifier
@@ -51,16 +51,17 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 81.82%
-- **Best Epoch**: 4
-- **Final Train Accuracy**: 86.37%
+- **Validation Accuracy**: 82.10%
+- **Best Epoch**: 5
+- **Final Train Accuracy**: 87.37%
 
 ### Per-Scale Performance
-- **Scale 256**: 81.82%
-- **Scale 512**: 81.94%
-- **Scale 768**: 81.97%
-- **Scale 1024**: 81.79%
-- **Scale 1280**: 81.45%
+- **Scale 256**: 82.10%
+- **Scale 512**: 82.34%
+- **Scale 768**: 82.28%
+- **Scale 1024**: 82.01%
+- **Scale 1280**: 81.90%
+- **Scale 1536**: 81.58%
 
 
 ## Usage
@@ -183,4 +184,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-10-12 05:13:59*
+*Generated on 2025-10-12 05:16:22*