Abstract:Introduction: Caricature is an exaggerated pictorial representation of a person, which is widely used in entertainment and political media. Recently, GAN-based methods achieved automatic caricature generation through transferring caricature style and performing shape exaggeration simultaneously. However, the caricature synthesized by these methods cannot perfectly reflect the characteristics of the subject, whose shape exaggeration are not reasonable and requires facial landmarks of caricature. In addition, the existing methods always produce the bad cases in caricature style due to the simpleness of their style transfer method. Methods: In this paper, we propose a Style Attention based Global-local Aware GAN to apply the characteristics of a subject to generate personalized caricature. To integrate the facial characteristics of a subject, we introduce a landmark-based warp controller for personalized shape exaggeration, which employs the facial landmarks as control points to warp image according to its facial features, without requirement of the facial landmarks of caricature. To fuse the facial feature with caricature style appropriately, we introduce a style-attention module, which adopts an attention mechanism, instead of the simple Adaptive Instance Normalization (AdaIN) for style transfer. To reduce the bad cases and increase the quality of generated caricatures, we propose a multi-scale discriminator to both globally and locally discriminate the synthesized and real caricature, which improves the whole structure and realistic details of the synthesized caricature. Results: Experimental results on two publicly available datasets, the WebCaricature and the CaVINet datasets, validate the effectiveness of our proposed method and suggest that our proposed method achieves better performance than the existing methods. Discussion: The caricatures generated by the proposed method can not only preserve the identity of input photo but also the characteristic shape exaggeration for each person, which are highly close to the real caricatures drawn by real artists. It indicates that our method can be adopted in the real application.

Semi-supervised Learning of Caricature Pattern from Manifold Regularization

Semi-Supervised Learning in Reconstructed Manifold Space for 3d Caricature Generation

Mapping learning in eigenspace for harmonious caricature generation.

Weakly-supervised caricature face parsing through domain adaptation

A Joint Local and Global Deep Metric Learning Method for Caricature Recognition

Learning to Caricature Via Semantic Shape Transform

A Regularized Approach for Geodesic-Based Semisupervised Multimanifold Learning

CariGAN: Caricature generation through weakly paired adversarial learning

A Semi-Supervised Face Recognition Method Based on Manifold Regularization

Unsupervised Domain Attention Adaptation Network for Caricature Attribute Recognition

Manifold regularization with GANs for semi-supervised learning

Style attention based global-local aware GAN for personalized facial caricature generation

AutoToon: Automatic Geometric Warping for Face Cartoon Generation

Controllable Facial Caricaturization with Localized Deformation and Personalized Semantic Attentions

3D-CariGAN: An End-to-End Solution to 3D Caricature Generation from Face Photos

Retrieval Based Cartoon Synthesis Via Heterogeneous Features Learning

Semi-Supervised Learning with GANs: Revisiting Manifold Regularization

Sample Based Face Caricature Generation

Face Manifold: Manifold Learning for Synthetic Face Generation

Manifold adversarial training for supervised and semi-supervised learning

3D-Carigan: an End-to-End Solution to 3D Caricature Generation from Normal Face Photos