Update README.md

README.md

@@ -4,60 +4,84 @@ tags:
 - deepfake-detection
 - computer-vision
 - efficientnet
+- xai
+- explainable-ai
 ---
 
 # XADE Deepfake Detector
 
-EfficientNet-B4 model trained for deepfake detection as part of the XADE (eXplainable Automated Deepfake Evaluation) thesis project.
+EfficientNet-B4 model trained for deepfake detection as part of the XADE
+(eXplainable Automated Deepfake Evaluation) thesis project at Jönköping
+University, 2026.
 
 ## Model Details
 
-- **Architecture:** EfficientNet-B4
+- **Architecture:** EfficientNet-B4 with custom two-layer classifier head
 - **Task:** Binary classification (real vs. fake faces)
-- **Training Dataset:** 140k Real and Fake Faces
-- **Test Accuracy:** 98.86%
-- **AUC-ROC:** 99.94%
+- **Training:** Progressive mixed training across 4 manipulation types
+- **Final checkpoint:** Run 4 (140k + CIPLAB + FF++ + Celeb-DF)
 
-## Performance
+## Cross-Dataset Performance (AUC-ROC)
 
-| Metric | Value |
-|--------|-------|
-| Accuracy | 98.86% |
-| Precision | 98.44% |
-| Recall | 99.28% |
-| F1-Score | 98.86% |
+| Dataset | Manipulation Type | AUC |
+|---|---|---|
+| 140k Real-Fake (training dist.) | GAN / StyleGAN synthesis | 0.9992 |
+| Fake-Vs-Real Hard | StyleGAN2 harder cases | 0.8948 |
+| FF++ derived | Neural face swap | 0.8789 |
+| CIPLAB | Photoshop manipulation | 0.7563 |
+| Celeb-DF v2 | High-quality face swap | 0.8049 |
+
+## Training Details
+
+- **Base dataset:** 140k Real and Fake Faces (StyleGAN-generated)
+- **Additional training data:** CIPLAB (~960 images), FF++ derived
+  (~1500 frames), Celeb-DF v2 (~1500 images)
+- **Training samples:** 100,000 per run (sampled from combined pool)
+- **Epochs:** 10 (early stopping patience 7)
+- **Optimizer:** AdamW with differential learning rates
+  (backbone: 1e-4, classifier: 1e-3)
+- **Batch size:** 64
+- **Validation accuracy:** 98.51%
+
+## Architecture
+```
+EfficientNet-B4 (ImageNet pretrained, last 30% unfrozen)
+└── Custom classifier head:
+    Dropout(0.5)
+    Linear(in_features → 512)
+    ReLU
+    BatchNorm1d(512)
+    Dropout(0.4)
+    Linear(512 → 2)
+```
 
 ## Usage
 ```python
 import torch
 from huggingface_hub import hf_hub_download
+from torchvision.models import efficientnet_b4
+import torch.nn as nn
 
 # Download model
 model_path = hf_hub_download(
-    repo_id="YOUR_USERNAME/xade-deepfake-detector",
+    repo_id="viktorahnstrom/xade-deepfake-detector",
     filename="best_model.pt"
 )
 
-# Load model
-checkpoint = torch.load(model_path)
-# ... (load into your model class)
+# Load checkpoint
+checkpoint = torch.load(model_path, map_location="cpu", weights_only=False)
+print(f"Trained for {checkpoint['epoch']} epochs")
+print(f"Classes: {checkpoint['class_names']}")  # ['fake', 'real']
 ```
 
-## Training Details
-
-- Samples: 100,000 training, 20,000 validation
-- Epochs: 10 (early stopping)
-- Optimizer: AdamW
-- Learning rate: 0.001
-- Batch size: 64
-
 ## Citation
 ```bibtex
 @misc{xade2026,
-  author = {Viktor Ahnström, Viktor Carlsson},
-  title = {XADE: Cross-Platform Explainable Deepfake Detection Using Vision-Language Models},
+  author = {Viktor Ahnström and Viktor Carlsson},
+  title = {XADE: Cross-Platform Explainable Deepfake Detection
+           Using Vision-Language Models},
   year = {2026},
-  publisher = {Hugging Face},
-  howpublished = {\url{https://huggingface.co/YOUR_USERNAME/xade-deepfake-detector}}
+  institution = {Jönköping University},
+  howpublished = {\url{https://huggingface.co/viktorahnstrom/xade-deepfake-detector}}
 }
 ```