Minor fix.
Browse files
README.md
CHANGED
|
@@ -5,7 +5,7 @@ language:
|
|
| 5 |
---
|
| 6 |
Pretrained model for Deepfake Video Detection Using Generative Convolutional Vision Transformer (GenConViT) paper.
|
| 7 |
|
| 8 |
-
|
| 9 |
|
| 10 |
The GenConViT model consists of two independent networks and incorporates the following modules:
|
| 11 |
|
|
@@ -14,7 +14,9 @@ The GenConViT model consists of two independent networks and incorporates the fo
|
|
| 14 |
ConvNeXt-Swin Hybrid layer
|
| 15 |
|
| 16 |
GenConViT is trained using Adam optimizer with a learning rate of 0.0001 and weight decay of 0.0001.
|
|
|
|
| 17 |
GenConViT is trained on the DFDC, FF++, and TM datasets.
|
|
|
|
| 18 |
GenConViT model has an average accuracy of 95.8% and an AUC value of 99.3% across the tested datasets (DFDC, FF++, and DeepfakeTIMT, Celeb-DF (v2)).
|
| 19 |
|
| 20 |
code link: https://github.com/erprogs/GenConViT
|
|
|
|
| 5 |
---
|
| 6 |
Pretrained model for Deepfake Video Detection Using Generative Convolutional Vision Transformer (GenConViT) paper.
|
| 7 |
|
| 8 |
+
GenConViT Model Architecture
|
| 9 |
|
| 10 |
The GenConViT model consists of two independent networks and incorporates the following modules:
|
| 11 |
|
|
|
|
| 14 |
ConvNeXt-Swin Hybrid layer
|
| 15 |
|
| 16 |
GenConViT is trained using Adam optimizer with a learning rate of 0.0001 and weight decay of 0.0001.
|
| 17 |
+
|
| 18 |
GenConViT is trained on the DFDC, FF++, and TM datasets.
|
| 19 |
+
|
| 20 |
GenConViT model has an average accuracy of 95.8% and an AUC value of 99.3% across the tested datasets (DFDC, FF++, and DeepfakeTIMT, Celeb-DF (v2)).
|
| 21 |
|
| 22 |
code link: https://github.com/erprogs/GenConViT
|