Deep lightweight face forgery detection network using multi-scale global features and adaptive weighted channel self-attention

Key Points

• Face forgery detection accuracy improved significantly through the use of multi-scale features and channel attention.
• The model achieved a detection rate of 95.5% on benchmark datasets, highlighting its efficacy across various scenarios.
• Employing a lightweight architecture, the detection network is suitable for real-time processing without sacrificing accuracy.
• These findings suggest a valuable tool for safeguarding against digital image manipulation, with broad applications in security.