Training-Free, Identity-Preserving Image Editing for Fashion Pose Alignment and Normalization

2026-07-14 · Source: cs.SE updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Retail Technology & Operations · Depth: Expert, extended

Summary

FashionRepose introduces a training-free, zero-shot pipeline for non-rigid pose editing of long-sleeve garments, specifically designed for the fashion industry. This solution addresses challenges in maintaining object identity and branding attributes during image transformations, which current diffusion models often struggle with. The pipeline integrates off-the-shelf models to adjust garment poses, enforcing a standardized 45-degree arm-to-torso alignment from an initial still-life configuration. It operates in near real-time, eliminating the need for specialized training data or fine-tuning. FashionRepose achieves consistent identity preservation, texture fidelity, and brand-specific attribute retention, offering a scalable solution for e-commerce, marketing, and design prototyping applications.

Key takeaway

For AI Engineers developing fashion e-commerce or marketing tools, FashionRepose offers a robust solution for automating garment pose normalization. You can achieve consistent, identity-preserving edits of long-sleeve clothing in near real-time without extensive model retraining. Consider integrating this training-free, multi-stage pipeline to enhance product visualization and streamline content creation, ensuring brand consistency across diverse digital platforms.

Key insights

FashionRepose offers a training-free, zero-shot pipeline for consistent, identity-preserving pose normalization of long-sleeve garments in fashion.

Principles

• Identity preservation is critical for brand consistency.
• Training-free pipelines enable real-time adaptability.
• Multi-stage integration of off-the-shelf models enhances precision.

Method

The pipeline involves long-sleeve detection, image preprocessing, coarse generation, conditioned unsampling, source-target shape matching, garment parts-composition, upsampling, and logo detection/suppression/injection.

In practice

• Automate garment repositioning for e-commerce.
• Generate consistent product images for marketing.
• Streamline design prototyping workflows.

Topics

• Fashion Image Editing
• Diffusion Models
• Pose Normalization
• Zero-Shot Learning
• Garment Repositioning
• Brand Identity Preservation

Best for: AI Scientist, Research Scientist, Machine Learning Engineer, Computer Vision Engineer, AI Engineer

Related on AIssential

• FASH-iCNN: Making Editorial Fashion Identity Inspectable Through Multimodal CNN Probing — FASH-iCNN reveals fashion's cultural logic by making editorial identity inspectable through multimodal CNN probing.
• Pro-Pose: Unpaired Full-Body Portrait Synthesis via Canonical UV Maps — Pro-Pose generates high-fidelity, reposed human avatars from single images by decoupling pose from texture in UV space.
• DetailAnywhere: Fashion Detail Generation via Cross-Modal Feature Alignment Distillation — DetailAnywhere generates photorealistic fashion details from a focus marker by aligning cross-modal features and optimizing consistency via reinforcement learning.
• Adaptive Identity Anchoring: Closed-Loop Keyframe Placement for Synthetic Paired Supervision in Video Face Swapping — Video face swapping can overcome identity drift and texture loss via adaptive anchoring and reality-referenced texture restoration.
• ReGenHuman: Re-Generating Human Appearances for Realistic Full-Body Video Anonymization — ReGenHuman pioneers realistic, temporally consistent full-body video anonymization by regenerating human appearances from structural cues.
• STiTch: Semantic Transition and Transportation in Collaboration for Training-Free Zero-Shot Composed Image Retrieval — Training-free zero-shot CIR improves by refining LLM-generated captions and using set-to-set alignment for diverse compositions.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.SE updates on arXiv.org.