A LOCAL-GLOBAL ADAPTER-BASED METHOD FOR SKETCH FACE RECOGNITION
DOI:
https://doi.org/10.62643/Abstract
Sketch-based face recognition aims to identify a person from a gallery of facial photographs using a sketch as the query input. However, large modality gaps between sketches and real images make direct matching highly unreliable. To address this challenge, we propose a Pix2Pix-based sketch face recognition framework that converts hand-drawn sketches into realistic face images before performing identification. The model employs a conditional GAN architecture to learn the mapping from sketch domain to photo domain, enabling the generation of structurally accurate and visually consistent facial images. The synthesized images are then fed into a standard face recognition network for feature extraction and matching. Our approach reduces the modality gap, enhances feature consistency across domains, and improves overall recognition accuracy. Experimental results on standard sketch–photo datasets demonstrate that the proposed method achieves significant improvements in identification performance compared to direct sketch-to-photo matching methods.
Downloads
Published
Issue
Section
License

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.













