Publications

Abstract

Face morphing is a powerful image processing technique that involves extracting facial landmarks from two images, aligning the faces, and blending their features using averaged landmarks. While traditional face morphing can distort individual identities, combining the FaceID model with the IP-Adapter presents an innovative approach to generating identity-preserving transformations.

FaceID extracts facial embeddings to retain key identity features, while the IP-Adapter adapts these embeddings to specific prompts, such as "older" or "younger," allowing for controlled and realistic transformations. The resulting solution has far-reaching applications, including predicting the future appearance of individuals, such as missing children, and offers a faster, more accessible alternative to conventional DNA testing methods.

By merging advanced facial recognition techniques with adaptable image models, this project achieves efficient, high-fidelity face morphing with identity preservation.

Process

Figure 1. Face Morphing Process - Extracting landmarks and blending faces.

Results

Figure 2. Resulting identity-preserving morphed face.