ReWeaver AI: 3D Garment Reconstruction from Sparse Images (CVPR 2026)

Meta Description: ReWeaver AI: the CVPR 2026 framework that reconstructs structured 3D garments and 2D sewing patterns from as few as 4 multi-view RGB images. How it works, applications, and where to try it.

Target Keyword: reweaver ai Secondary Keywords: reweaver AI garment reconstruction, reweaver CVPR 2026, reweaver 3D garment, reweaver sewing pattern

What Is ReWeaver?

ReWeaver is a research framework presented at CVPR 2026 that reconstructs structured 3D garment geometry and 2D sewing patterns from sparse multi-view RGB images. Given as few as 4 photographs of a garment from different angles, ReWeaver can generate a simulation-ready 3D model with accurate topology and a corresponding 2D sewing pattern.

This is a significant advancement for the fashion, gaming, and virtual try-on industries, where creating realistic digital garments traditionally requires manual 3D modeling or expensive 3D scanning equipment.

Paper: “ReWeaver: Towards Simulation-Ready and Topology-Accurate Garment Reconstruction” Conference: CVPR 2026 Code: github.com/SII-LiMing/ReWeaver-Code

How ReWeaver Works

The Problem

Existing garment reconstruction methods either:

• Produce visually plausible but topologically incorrect meshes (can’t be simulated)
• Require dense input (hundreds of images or 3D scans)
• Can’t generate sewing patterns (needed for physical fabrication)

ReWeaver’s Approach

ReWeaver uses a dual-branch architecture:

Branch 1: 3D Geometry Reconstruction

• Takes sparse multi-view RGB images (minimum 4)
• Reconstructs the 3D garment mesh with accurate topology
• Output is simulation-ready (can be used in cloth simulation engines)

Branch 2: 2D Sewing Pattern Generation

• Generates the flat 2D sewing pattern pieces from the 3D reconstruction
• Patterns are physically accurate and can be used for actual garment fabrication
• Captures seam lines, darts, and panel boundaries

Key Innovation

The breakthrough is joint optimization: the 3D geometry and 2D sewing pattern are reconstructed simultaneously, ensuring they’re mutually consistent. Previous methods treated these as separate problems, leading to mismatches between the 3D model and the pattern.

Applications

Technical Specifications

How to Try ReWeaver

1. Visit the GitHub repository: github.com/SII-LiMing/ReWeaver-Code
2. Clone the repository and install dependencies
3. Prepare 4+ multi-view RGB images of a garment
4. Run the reconstruction pipeline
5. Export 3D mesh and 2D sewing pattern

FAQ

What is ReWeaver AI?

ReWeaver is a CVPR 2026 research framework that reconstructs 3D garment geometry and 2D sewing patterns from sparse multi-view RGB images. It produces simulation-ready meshes and physically accurate patterns.

How many images does ReWeaver need?

Minimum 4 multi-view RGB images of the garment from different angles.

Can ReWeaver be used commercially?

The code is released for research use. Check the license terms on the GitHub repository for commercial applications.

What’s the difference between ReWeaver and other 3D garment tools?

ReWeaver uniquely produces both 3D geometry and 2D sewing patterns simultaneously, ensuring they’re mutually consistent. Other tools typically produce one or the other, or treat them as separate problems.

Conclusion

ReWeaver represents a meaningful step forward in digital garment reconstruction. By jointly optimizing 3D geometry and 2D sewing patterns from just 4 images, it opens new possibilities for fashion tech, e-commerce, and digital content creation.

For AI-generated video content featuring fashion and apparel, explore the VideosPrompt community for garment-related video prompts.

Last updated: September 2026

Code: github.com/SII-LiMing/ReWeaver-Code