Advances in Human-Centric Image and Video Processing

The field of human-centric image and video processing is rapidly advancing, with a focus on developing more accurate and robust methods for tasks such as face super-resolution, human pose estimation, and facial age editing. A key trend in this area is the use of diffusion-based models, which have been shown to achieve state-of-the-art performance in a range of applications. Another important direction is the development of methods that can preserve identity and other important attributes, such as pose and clothing, while editing or manipulating human images. Notable papers in this area include: Personalized Face Super-Resolution with Identity Decoupling and Fitting, which proposes a novel method for face super-resolution that enhances ID restoration under large scaling factors. TimeMachine, a diffusion-based framework that achieves accurate age editing while keeping identity features unchanged. GANDiff FR, a synthetic framework that precisely controls demographic and environmental factors to measure, explain, and reduce bias in face recognition. Odo, an end-to-end diffusion-based method that enables realistic and intuitive body reshaping guided by simple semantic attributes.

Sources

Personalized Face Super-Resolution with Identity Decoupling and Fitting

A Coarse-to-Fine Human Pose Estimation Method based on Two-stage Distillation and Progressive Graph Neural Network

TimeMachine: Fine-Grained Facial Age Editing with Identity Preservation

GANDiff FR: Hybrid GAN Diffusion Synthesis for Causal Bias Attribution in Face Recognition

Stable Diffusion-Based Approach for Human De-Occlusion

Odo: Depth-Guided Diffusion for Identity-Preserving Body Reshaping

HyperDiff: Hypergraph Guided Diffusion Model for 3D Human Pose Estimation

Built with on top of