Update README.md

README.md

@@ -1,4 +1,15 @@
- Vision Transformer (ViT) Models for Deepfake Detection
+---
+license: apache-2.0
+datasets:
+- maxin-cn/FaceForensics
+metrics:
+- accuracy
+base_model:
+- google/vit-base-patch16-224
+tags:
+- Forensics
+---
+## Vision Transformer (ViT) Models for Deepfake Detection
 
 This repository provides Vision Transformer (ViT) models fine-tuned to detect manipulated (fake) versus authentic (real) image frames extracted from the DeepFake Detection (DFD) dataset. The models were trained using Intel® ARC GPU (XPU-enabled) and optimized for binary image classification in digital forensics workflows.
 
@@ -15,7 +26,6 @@ This repository provides Vision Transformer (ViT) models fine-tuned to detect ma
 | `vit_large_patch32_224`   | ImageNet-21k          | 32×32      | ~304M      |
 | `vit_huge_patch14_224`    | ImageNet-21k          | 14×14      | ~632M      |
 
-> ✅ All models are fine-tuned with `num_classes=2` for binary classification (`real` vs `fake`).
 
 ---
 
@@ -76,5 +86,4 @@ This repository is intended for research and educational purposes only.
 
 ## 🙋 Author
 
-Trained and curated by an Intel-XPU researcher working on multimodal and digital forensics pipelines.  
-
+Trained and curated by an Intel-XPU researcher working on multimodal and digital forensics pipelines.