Deepfake Image Detector (CILab Fine-Tuned)

Model Performance

  • Test Accuracy: 73.17%
  • Best Validation Accuracy: 86.27%
  • Best Epoch: 11
  • Planned Epochs: 12
  • Actual Epochs Trained: 12 (early stopping applied)

Dataset

  • Training: 1,632 images
  • Validation: 204 images
  • Test: 205 images

Training Details

Training was stopped early at epoch 12 due to early stopping criteria being met. The best model was achieved at epoch 11 with validation accuracy of 86.27%.

Usage

from transformers import ViTForImageClassification, ViTFeatureExtractor

model = ViTForImageClassification.from_pretrained('shivani1511/deepfake-image-detector-new-latest')
feature_extractor = ViTFeatureExtractor.from_pretrained('shivani1511/deepfake-image-detector-new-latest')

Notes

  • Fine-tuned on CILab dataset to address AI-generated fake detection (e.g., StyleGAN).
  • Base model: shivani1511/deepfake-image-detector-new (Vision Transformer).
  • Improvements: Enhanced data augmentation, class-weighted loss, Mixup, more unfrozen layers.
Downloads last month
5
Safetensors
Model size
85.8M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support