Jointly Conditioned Diffusion Model for Multi-View Pose-Guided Person Image Synthesis

2025-11-19

AI Summary

To address texture incompleteness and poor cross-view consistency in pose-guided human image generation from single-view reference images, this paper proposes a joint conditional diffusion model. Methodologically, it introduces (1) an Appearance Prior Module (APM) that explicitly models multi-view correspondences of identity, color, and texture across poses; and (2) a Joint Conditional Injection (JCI) mechanism that adaptively fuses multi-view features and injects them into the denoising network—supporting variable numbers of reference inputs while preserving architectural simplicity. Evaluated on standard benchmarks, the method significantly improves visual fidelity and cross-view geometric-appearance consistency of generated images, achieving state-of-the-art performance. It further demonstrates strong generalization capability across diverse poses, identities, and clothing styles.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Pose-guided human image generation is limited by incomplete textures from single reference views and the absence of explicit cross-view interaction. We present jointly conditioned diffusion model (JCDM), a jointly conditioned diffusion framework that exploits multi-view priors. The appearance prior module (APM) infers a holistic identity preserving prior from incomplete references, and the joint conditional injection (JCI) mechanism fuses multi-view cues and injects shared conditioning into the denoising backbone to align identity, color, and texture across poses. JCDM supports a variable number of reference views and integrates with standard diffusion backbones with minimal and targeted architectural modifications. Experiments demonstrate state of the art fidelity and cross-view consistency.

Problem

• Generating person images from incomplete single-view texture references
• Addressing lack of explicit cross-view interaction in pose-guided synthesis
• Aligning identity, color and texture consistency across multiple poses

Innovation

• Jointly conditioned diffusion framework with multi-view priors
• Appearance prior module infers holistic identity from references
• Joint conditional injection fuses multi-view cues for alignment