Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

A two-level generative model for cloth representation and shape from shading.

Feng Han1, Song-Chun Zhu

  • 1Department of Computer Science and Statistics, University of California, Los Angeles, CA 90095, USA. hanf@cs.ucla.edu

IEEE Transactions on Pattern Analysis and Machine Intelligence
|May 15, 2007
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Snapshot polarized light scattering spectroscopy using spectrally-modulated polarimetry for early gastric cancer detection.

Journal of biophotonics·2021
Same author

Fatty Acid-Binding Proteins Aggravate Cerebral Ischemia-Reperfusion Injury in Mice.

Biomedicines·2021
Same author

Effect of <i>N</i>, <i>N</i>-Dimethylglycine on Homocysteine Metabolism in Rats Fed Folate-Sufficient and Folate-Deficient Diets.

Biomedical and environmental sciences : BES·2021
Same author

Reduced coupling between cerebrospinal fluid flow and global brain activity is linked to Alzheimer disease-related pathology.

PLoS biology·2021
Same author

Efficient Construction of 5H-1,4-Benzodiazepine Derivatives by a Catalyst-Free Direct Aerobic Oxidative Annulation Strategy.

ChemSusChem·2021
Same author

An Effective and Efficient Sample Preparation Method for 2-Methyl-Isoborneol and Geosmin in Fish and Their Analysis by Gas Chromatography-Mass Spectrometry.

International journal of analytical chemistry·2021
Same journal

HardFlow: Hard-Constrained Sampling for Flow-Matching Models Via Trajectory Optimization.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Industrial Brain: Self-Evolving Neuro-Symbolic Autonomy with Causal Resilience for Cyber-Physical Systems.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Adaptive Hardness-Driven Dictionary Distillation for Incomplete Streaming View Clustering.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Mixture of Global and Local Experts with Diffusion Transformer for Controllable Face Generation.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Achieving Text-based Person Retrieval with Any Granularity.

IEEE transactions on pattern analysis and machine intelligence·2026
See all related articles

This study introduces a novel two-level generative model for accurately reconstructing 3D cloth shapes from images. The model improves shape-from-shading by using learned primitives for folds and a Markov random field for flat areas.

Area of Science:

  • Computer Vision
  • Computer Graphics
  • Machine Learning

Background:

  • Reconstructing 3D surface geometry from 2D images is a fundamental challenge in computer vision.
  • Traditional shape-from-shading (SFS) methods often struggle with accuracy due to the ill-posed nature of the problem.
  • Representing complex surfaces like drapery requires sophisticated models that capture both fine details and overall structure.

Purpose of the Study:

  • To develop a robust two-level generative model for representing drapery and clothing images and their corresponding surface depth maps.
  • To enhance the performance of shape-from-shading by incorporating middle-level visual knowledge in the form of learned primitives.
  • To provide a more accurate and stable method for 3D surface reconstruction of deformable objects.

Main Methods:

Related Experiment Videos

  • A two-level generative model is proposed, with the upper level handling high-contrast fold areas using learned shading and fold primitives.
  • The lower level reconstructs flat areas using a smoothness prior (Markov random field).
  • A supervised learning phase using photometric stereo data trains the parametric primitives. The method involves inferring folds, estimating 3D fold structures, and then applying conventional SFS with boundary conditions.

Main Results:

  • The proposed two-level model significantly improves upon classical shape-from-shading by reducing dimensionality and integrating primitive-based knowledge.
  • Experiments demonstrate more robust 3D surface reconstruction results compared to state-of-the-art methods.
  • The model effectively captures the complex geometry of drapery and clothing surfaces.

Conclusions:

  • The two-level generative model offers a powerful approach for shape-from-shading and general shape-from-X problems.
  • This representation can be viewed as a two-level inhomogeneous Markov random field model.
  • The study revisits Marr's concept of computing a 2(1/2)D sketch from a primal sketch, offering a new perspective on 3D reconstruction.