NaifA1i/attention-enhanced-deepfake-detection: Attention-Enhanced Deepfake Detection (v1.0.0)

Overview

This release contains the implementation of the proposed deepfake detection model based on ResNet50V2 with SE and CBAM attention mechanisms.

Pipeline

1. Face Extraction (MyCodeForExtractFaceFromVideo.py)
2. Normalization (Normalize.py)
3. Data Quality Filtering (DataQuality.py)
4. Model Training & Evaluation (MainCode.py)

Requirements

See requirements.txt

Usage

Run scripts sequentially: python MyCodeForExtractFaceFromVideo.py python Normalize.py python DataQuality.py python MainCode.py

Citation

If you use this work, please cite:

Attention-Enhanced Deepfake Detection Using ResNet50V2 and Hybrid Attention Mechanisms
Naif A. Almalki, Mahmoud Ragab, Faris Kateb
Under review at The Visual Computer (Springer), 2026.