Face2QR: A Unified Framework for Aesthetic, Face-Preserving, and Scannable QR Code Generation

TL;DR

Face2QR introduces a training-free pipeline that balances face ID preservation, aesthetics, and QR decodability by combining three modules: IDQR to unify style with face identity, IDRS to reconcile face ID with QR patterns via a blueprint and module reshuffling, and IDSE to enhance scannability through latent-code optimization. The approach uses a unified Stable Diffusion framework with dual control networks and a Latent Diffusion-based refinement that preserves facial features while maintaining robust decoding through a spatially dynamic loss. Empirical results show state-of-the-art performance in preserving face identity and aesthetics, with high scanning robustness across sizes and angles, and strong subjective preferences in user studies. The work enables personalized, visually appealing QR codes suitable for social media and marketing, while acknowledging limitations in computation and model stability and discussing broader societal implications and ethics.

Abstract

Existing methods to generate aesthetic QR codes, such as image and style transfer techniques, tend to compromise either the visual appeal or the scannability of QR codes when they incorporate human face identity. Addressing these imperfections, we present Face2QR-a novel pipeline specifically designed for generating personalized QR codes that harmoniously blend aesthetics, face identity, and scannability. Our pipeline introduces three innovative components. First, the ID-refined QR integration (IDQR) seamlessly intertwines the background styling with face ID, utilizing a unified Stable Diffusion (SD)-based framework with control networks. Second, the ID-aware QR ReShuffle (IDRS) effectively rectifies the conflicts between face IDs and QR patterns, rearranging QR modules to maintain the integrity of facial features without compromising scannability. Lastly, the ID-preserved Scannability Enhancement (IDSE) markedly boosts scanning robustness through latent code optimization, striking a delicate balance between face ID, aesthetic quality and QR functionality. In comprehensive experiments, Face2QR demonstrates remarkable performance, outperforming existing approaches, particularly in preserving facial recognition features within custom QR code designs. Codes are available at $\href{https://github.com/cavosamir/Face2QR}{\text{this URL link}}$.

Figures (6)

• Figure 1: Face images (first row) and QR code images (second row) generated by Face2QR. Our QR codes not only faithfully maintain face ID, but also showcase remarkable scanning resilience and aesthetic quality.
• Figure 2: The pipeline of Face2QR is a training-free process for generating ID-consistent and scannable QR code images. Our pipeline has three stages, represented by blue, red, and green arrows. The IDRS module resolves conflicts between human identity and QR patterns during the control process, while the IDSE module reduces coding errors to ensure the output is scannable.
• Figure 3: Illustration of IDRS (left) and IDSE (right). In IDRS, we maintain the information codes within the face and marker regions (red and yellow masks) and remap the remaining modules accordingly. In IDSE, we strengthen the finder and alignment pattern, and update in latent space using adaptive loss to enhance scannability. Visualization $D$ shows the difference between $I^o$ and $\hat{I^s}$. Compared with uniform loss, adaptive loss modifies face region more gently.
• Figure 4: User study of different methods.
• Figure 5: Curve of different metrics during IDSE. We show metric curves for diverse samples, each represented by distinct colors. These curves illustrate metric variations over 300 iterations.