# xdfdet: Explainable Deepfake Detection | Eschatia Labs

Source: https://eschatialabs.com/projects/xdfdet/

Project · computer vision · open source

Explainable deepfake detection

Eschatia Labs built xdfdet to check each frame of a video for AI-altered faces. It uses EfficientNet-B4, an image recognition model, and Grad-CAM heat maps that show which parts of the face each decision relied on. Its best model reaches 0.897 AUC on FaceForensics++, where 0.5 is guessing and 1 is perfect.

[GitHub](https://github.com/mertkayacs/xdfdet) [Models](https://huggingface.co/mertkayacs/xdfdet) [Play: real or AI?](https://xdfdet.mertkayacs.com/game/)

**0.897**AUC on FaceForensics++

**8**trained models released

**9**training methods compared

**UBMK 2026**accepted

## What it does

xdfdet sorts video frames into real and altered and explains each decision with a Grad-CAM heat map over the face. The study behind it compares nine training methods, including augmentation, which varies the training images, and cutout, which blacks out parts of them. It measures accuracy, how closely confidence matches results, and which parts of the face each model relies on.

On FaceForensics++, a public collection of real and altered face videos, EfficientNet-B4 reaches 0.897 AUC with varied and partly blacked-out training images, versus 0.868 without them. The repository holds the training and evaluation code and eight trained models.

## Materials

Eight EfficientNet-B4 models on Hugging Face and Kaggle, the training and evaluation code, the UBMK 2026 paper and the MSc thesis it grew from.

[github.com/mertkayacs/xdfdet](https://github.com/mertkayacs/xdfdet) [huggingface.co/mertkayacs/xdfdet](https://huggingface.co/mertkayacs/xdfdet) [doi.org/10.5281/zenodo.18998566](https://doi.org/10.5281/zenodo.18998566) [xdfdet.mertkayacs.com (demo game)](https://xdfdet.mertkayacs.com/game/)

![Eschatia Labs](https://eschatialabs.com/art/prototype/wordmark-pixel.png)

At the edge of the map,
where the old charts end.

[Eschatia Labs](https://eschatialabs.com/), an independent, non-profit open-source AI research lab founded by [Mert Kaya](https://mertkayacs.com/) and [Betül Keleş Kaya](https://eschatialabs.com/researchers/#betul-keles-kaya).
[mertkayacs@gmail.com](mailto:mertkayacs@gmail.com)
Last updated: October 25, 2026
Eschatia Labs · 2026

## Eschatia Labs

- [Research](https://eschatialabs.com/research/)
- [Researchers](https://eschatialabs.com/researchers/)
- [Benchmarks](https://eschatialabs.com/benchmarks/)
- [xdfdet](https://eschatialabs.com/projects/xdfdet/)
- [About](https://eschatialabs.com/about/)

## Projects

- [Emberwick](https://emberwick.mertkayacs.com/)
- [reevesagents](https://reevesagents.mertkayacs.com/)
- [JevAlt](https://jevalt.mertkayacs.com/)
- [xdfdet](https://xdfdet.mertkayacs.com/)
- [Tholos](https://tholos.mertkayacs.com/)
- [JevOss](https://jevoss.mertkayacs.com/)
- [Eldalambë](https://eldalambe.mertkayacs.com/)

## Elsewhere

- [Mert Kaya](https://mertkayacs.com/)
- [GitHub](https://github.com/mertkayacs)
- [Hugging Face](https://huggingface.co/mertkayacs)
- [Kaggle](https://www.kaggle.com/mertilovski)
- [ORCID](https://orcid.org/0009-0006-7018-3027)

- [English](https://eschatialabs.com/projects/xdfdet/)
- [Türkçe](https://eschatialabs.com/tr/projects/xdfdet/)
- [Deutsch](https://eschatialabs.com/de/projects/xdfdet/)
- [简体中文](https://eschatialabs.com/zh/projects/xdfdet/)
- [Italiano](https://eschatialabs.com/it/projects/xdfdet/)
- [Español](https://eschatialabs.com/es/projects/xdfdet/)

[Image credits](https://eschatialabs.com/about/#credits) · [Back to top](https://eschatialabs.com/projects/xdfdet/#main)