Cross multiscale vision transformer for deep fake detection

TL;DR

This paper evaluates deep fake detection methods using the SP Cup 2025 dataset, exploring various deep learning architectures to improve detection accuracy and robustness against manipulated media.

Contribution

It introduces a comprehensive evaluation of multiple deep learning models, including a novel cross multiscale vision transformer, for deep fake detection.

Findings

Transformers outperform traditional CNNs in detection accuracy

Multiscale models show improved robustness against various deep fake techniques

Achieved state-of-the-art results on the SP Cup 2025 dataset

Abstract

The proliferation of deep fake technology poses significant challenges to digital media authenticity, necessitating robust detection mechanisms. This project evaluates deep fake detection using the SP Cup's 2025 deep fake detection challenge dataset. We focused on exploring various deep learning models for detecting deep fake content, utilizing traditional deep learning techniques alongside newer architectures. Our approach involved training a series of models and rigorously assessing their performance using metrics such as accuracy.