Explainable deepfake detection - a mertkayacs Collection

mertkayacs 's Collections

Eschatia Labs: open models and data

JevAlt decision models

Small local LLMs in GGUF

Tholos: local AI agents

Explainable deepfake detection

Turkish and German AI models

Benchmarks and datasets

Explainable deepfake detection

updated 4 days ago

EfficientNet-B4 deepfake detectors with Grad-CAM facial-region analysis, FaceForensics++ results and research limits. Non-commercial weights.

mertkayacs/xdfdet

Video Classification • Updated 1 day ago • 698 • 1

Note Eight EfficientNet-B4 detectors with Grad-CAM. Best released AUC: 0.8981 on FaceForensics++; splits differ. Non-commercial weights.
FaceForensics++: Learning to Detect Manipulated Facial Images

Paper • 1901.08971 • Published Jan 25, 2019 • 1

Note The deepfake video dataset the detectors were trained and scored on.
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks

Paper • 1905.11946 • Published May 28, 2019 • 4

Note The image backbone behind every xdfdet detector (EfficientNet-B4).
Improved Regularization of Convolutional Neural Networks with Cutout

Paper • 1708.04552 • Published Aug 15, 2017

Note The masking method behind the cutout training variants.
Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization

Paper • 1610.02391 • Published Oct 7, 2016

Note The explanation method used to map the facial regions behind each prediction.