Update README.md

README.md

@@ -56,17 +56,20 @@ They are well-suited for deployment in digital forensics, content moderation, an
 
 ---
 
-## 📊 Results
-
-| Vision Transformer Model       | Train Accuracy | Validation Accuracy |
-|-------------------------------|----------------|---------------------|
-| **vit_large_patch16_224**     | 94.89%         | **91.22%**          |
-| vit_large_patch32_224         | 91.31%         | 89.23%              |
-| vit_tiny_patch16_224          | 92.41%         | 89.20%              |
-| vit_small_patch32_224         | 91.58%         | 88.38%              |
-| vit_small_patch16_224         | 84.72%         | 87.68%              |
-| vit_base_patch16_224          | 90.65%         | 85.36%              |
-| vit_base_patch32_224          | 79.54%         | 79.54%              |
+## 📊 Results 
+
+| Model | Train Accuracy | Validation Accuracy |
+|-------|----------------|---------------------|
+| [vit_large_patch16_224](https://huggingface.co/pranav2711/VisionTransformerDigitalForensics/blob/main/vit_large_patch16_224.pth) | **94.89%** | **91.22%** |
+| [vit_large_patch32_224](https://huggingface.co/pranav2711/VisionTransformerDigitalForensics/blob/main/vit_large_patch32_224.pth) | 91.31% | 89.23% |
+| [vit_tiny_patch16_224](https://huggingface.co/pranav2711/VisionTransformerDigitalForensics/blob/main/vit_tiny_patch16_224.pth) | 92.41% | 89.20% |
+| [vit_small_patch32_224](https://huggingface.co/pranav2711/VisionTransformerDigitalForensics/blob/main/vit_small_patch32_224.pth) | 91.38% | 88.29% |
+| [vit_small_patch16_224](https://huggingface.co/pranav2711/VisionTransformerDigitalForensics/blob/main/vit_small_patch32_224.pth) | 80.67% | 81.25% |
+| [vit_base_patch16_224](http://huggingface.co/pranav2711/VisionTransformerDigitalForensics/blob/main/vit_base_patch16_224.pth) | 90.65% | 85.36% |
+| [vit_base_patch32_224](https://huggingface.co/pranav2711/VisionTransformerDigitalForensics/blob/main/vit_base_patch32_224.pth) | 79.54% | 79.54% |
+
+**vit_large_patch16_224**
+This model achieved the highest validation accuracy of 91.22% with strong training stability and generalization. It is recommended as the final model for deployment or downstream tasks.
 
 ---