Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
126 tokens/sec
GPT-4o
47 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Garment Recovery with Shape and Deformation Priors (2311.10356v2)

Published 17 Nov 2023 in cs.CV

Abstract: While modeling people wearing tight-fitting clothing has made great strides in recent years, loose-fitting clothing remains a challenge. We propose a method that delivers realistic garment models from real-world images, regardless of garment shape or deformation. To this end, we introduce a fitting approach that utilizes shape and deformation priors learned from synthetic data to accurately capture garment shapes and deformations, including large ones. Not only does our approach recover the garment geometry accurately, it also yields models that can be directly used by downstream applications such as animation and simulation.

Citations (6)

Summary

  • The paper presents a novel two-stage fitting process that first optimizes a pre-trained deformation model and then refines mesh details to capture loose-fitting garment shapes.
  • It leverages both 2D and 3D parameterization through the Implicit Sewing Patterns (ISP) model to represent complex garment deformations.
  • The method outperforms state-of-the-art approaches in Chamfer Distance and IoU metrics, demonstrating superior reconstruction accuracy and visual realism.

An Overview of "Garment Recovery with Shape and Deformation Priors"

The paper "Garment Recovery with Shape and Deformation Priors" introduces a sophisticated method for generating realistic 3D models of garments. While significant progress has been made in modeling tight-fitting clothing, loose-fitting garments continue to present a considerable challenge in the reconstruction of clothed humans. This research addresses this gap by proposing a novel method that effectively recovers the shape of loose-fitting garments from single images, leveraging shape and deformation priors learned from synthetic data.

Methodology

The approach centers around a novel fitting process that integrates both shape and deformation priors, utilizing the Implicit Sewing Patterns (ISP) model as a foundation. ISP employs 2D and 3D parameterization to represent garments based on a set of individual 2D panels and associated 3D surfaces. This enables the model to capture a wide array of garment deformations, including those that deviate significantly from the body shape. The method involves a two-stage fitting process: First, optimizing the parameters of a pre-trained deformation model to produce an initial garment shape and then refining the mesh to capture fine local details.

The model begins by determining vertex positions using ISP, generating a prototype garment that initially fits closely to the underlying body. Subsequently, a deformation model, driven by pixel-aligned image features derived from estimated garment normals and body segmentation, predicts any necessary corrective displacements to accommodate larger deformations characteristic of loose-fitting garments.

Numerical Results and Comparisons

The paper presents strong quantitative results, indicating that the proposed method outperforms existing state-of-the-art approaches in both reconstruction accuracy and visual realism, particularly for loose-fitting and free-flowing garments such as skirts and open jackets. Specifically, the method highlights its superiority in terms of Chamfer Distance (CD) and Intersection over Union (IoU) metrics when compared against methods like SMPLicit, DrapeNet, and others. This improved performance is attributed to the model's ability to leverage detailed deformation priors, which are fine-tuned for each image to enhance alignment with real-world observations.

Implications and Future Directions

The introduction of this method has significant implications for several applications, including fashion design, virtual try-on, and augmented reality, where the realistic reconstruction of clothing is crucial. This work not only bridges the gap in reconstructing loose-fitting garments but also provides a scalable framework for enhancing garment modeling in varied contexts.

Looking forward, there is potential for further research in extending this model to handle temporal sequences, enhancing deformation over time to achieve consistent reconstruction in videos. Such advancements could facilitate more robust animation and simulation applications, expanding the practical utility of the model in the field of computer graphics and virtual environments.

In conclusion, "Garment Recovery with Shape and Deformation Priors" presents a robust solution to a challenging problem in garment modeling and sets a foundation for future developments in realistic 3D garment reconstruction. The research offers a substantial contribution to both theoretical modeling techniques and practical applications in garment-based immersive technologies.

X Twitter Logo Streamline Icon: https://streamlinehq.com
Youtube Logo Streamline Icon: https://streamlinehq.com