What AI says about DeepfakeDetection

DeepfakeDetection is a project that detects deepfake images using modern CNN/ViT backbones (EfficientNet, FasterViT, EfficientFormerV2-S1) to balance speed and accuracy. Models are integrated via timm, enabling reproducible training/evaluation and Grad-CAM visualizations for interpretability. The approach diversifies architectures to reduce brittleness to dataset-specific artifacts, and the repo includes documentation and a linked research paper.