Researchers introduce CASIA FaceSwapping benchmark and comprehensive survey organizing face-swapping methods into five paradigms

A new survey paper presents a structured review of face-swapping methods, organizing them into five major paradigms and analyzing their design principles, strengths, and limitations.

The authors introduce CASIA FaceSwapping, a benchmark dataset with balanced demographic distributions and explicit attribute variations, along with standardized protocols to evaluate the robustness of face-swapping approaches.

The work addresses fragmentation in the field: existing face-swapping methods remain scattered across different approaches, and their evaluation has lacked consistency due to the absence of standardized datasets and evaluation protocols.