DAC-Pose: Dual-Agent Collaborative Framework for Pose-Guided Human Generation

2026-08-05

AI Summary

This work addresses the challenge of generating realistic human images under large pose variations, where existing pose-guided methods often produce visual artifacts due to difficulties in inferring occluded regions and modeling complex deformations. To overcome this limitation, the authors propose a dual-agent collaborative generation framework that introduces, for the first time, a complementary mechanism between semantic reasoning and visual perception. The Prior Semantic Reasoning (PSR) agent performs fine-grained attribute inference, while the Discrepancy-Aware Visual Encoding (DAVE) agent explicitly models view-induced geometric misalignment and provides spatial constraints. These two agents interact through an autonomous feedback loop to jointly optimize the synthesis process. Extensive experiments on DeepFashion and Market-1501 demonstrate that the proposed method significantly outperforms state-of-the-art approaches, effectively preserving texture alignment and identity consistency even under extreme viewpoint changes.

Abstract

AI agents have emerged as a powerful new paradigm in generative image synthesis, enabling systems to perform complex semantic reasoning rather than passive pixel-level mapping. In pose-guided human generation, conventional methods inevitably produce severe visual artifacts under drastic viewpoint shifts, fundamentally because they lack the cognitive capacity to logically deduce unseen regions and model complex spatial deformations. To bridge this gap, we propose DAC-Pose, a novel agent-driven multimodal framework that reformulates single-view human generation as a collaborative dual-agent system. DAC-Pose integrates two complementary components, namely, the Prior Semantic Reasoning (PSR) agent and the Discrepancy-Aware Visual Encoding (DAVE) agent. Functioning as a cognitive engine, PSR utilizes collaborative reasoning to deduce the fine-grained attributes of unseen regions. Concurrently, acting as a specialized visual perception agent, DAVE quantifies and encodes viewpoint-induced spatial misalignments, continuously feeding robust spatial constraints back into the generative process. This autonomous feedback loop between semantic deduction and visual perception ensures high-fidelity detail synthesis. Extensive experiments on the DeepFashion and Market-1501 benchmarks validate the superiority of our agent-driven paradigm. Notably, DAC-Pose excels in preserving texture alignment and identity consistency under drastic viewpoint changes. The code is available at https://github.com/AIVRC/DAC-Pose.

Problem

• pose-guided human generation
• viewpoint shifts
• visual artifacts
• spatial deformations
• unseen regions

Innovation

• dual-agent collaboration
• pose-guided human generation
• semantic reasoning
• spatial deformation modeling
• viewpoint-invariant synthesis