Advances in Image and Face Representation

The field of computer vision is moving towards more sophisticated and nuanced representations of images and faces. Researchers are exploring new methods for facial stylization, face editing, and image representation, with a focus on preserving semantic information and achieving high-fidelity results. One notable trend is the use of Gaussian-based approaches, such as Gaussian Splatting and HyperGaussians, which offer improved expressivity and efficiency. Another area of research is the development of more effective and efficient methods for virtual try-on, outfit retrieval, and recommendation. Overall, the field is advancing rapidly, with new techniques and architectures being proposed to address ongoing challenges in image and face representation. Notable papers include: Instant GaussianImage, which proposes a generalizable and self-adaptive image representation framework, and HyperGaussians, which introduces a novel extension of 3D Gaussian Splatting for high-quality animatable face avatars. DiffFit is also noteworthy, as it achieves high-fidelity virtual try-on through a two-stage latent diffusion framework.

Sources

Advancing Facial Stylization through Semantic Preservation Constraint and Pseudo-Paired Supervision

SemFaceEdit: Semantic Face Editing on Generative Radiance Manifolds

Attention to Burstiness: Low-Rank Bilinear Prompt Tuning

DiffFit: Disentangled Garment Warping and Texture Refinement for Virtual Try-On

KiseKloset: Comprehensive System For Outfit Retrieval, Recommendation, And Try-On

Instant GaussianImage: A Generalizable and Self-Adaptive Image Representation via 2D Gaussian Splatting

GViT: Representing Images as Gaussians for Visual Recognition

HiNeuS: High-fidelity Neural Surface Mitigating Low-texture and Reflective Ambiguity

ViscoReg: Neural Signed Distance Functions via Viscosity Solutions

FreNBRDF: A Frequency-Rectified Neural Material Representation

AvatarMakeup: Realistic Makeup Transfer for 3D Animatable Head Avatars

HyperGaussians: High-Dimensional Gaussian Splatting for High-Fidelity Animatable Face Avatars

Built with on top of