Diff-Palm: Realistic Palmprint Generation with Polynomial Creases and Intra-Class Variation Controllable Diffusion Models

Figure 1. Comparison between PCE-Palm [20] and the proposed Diff-Palm. (a) PCE-Palm adopts conditional GAN with Bézier creases [44] as input to generate palmprint datasets. Diff-Palm introduces a polynomial crease and a novel diffusion model for synthesizing datasets with adjustable intra-class variations. (b) The average performance of recognition models, trained on three types of datasets (real data, PCE-Palm generated, and Diff-Palm generated) and evaluated on five public datasets.