Skip to content

← ArchivePaper2016

AutoHair: Fully Automatic Hair Modeling from a Single Image

Menglei Chai, Tianjia Shao, Hongzhi Wu, Yanlin Weng, Kun Zhou

SIGGRAPHAcademic99 cites9 descendantsCFXML Deformation

First fully automatic single-image hair modeling method using a hierarchical deep network for segmentation and direction estimation, building a 50K-model hairstyle database.

Abstract

We introduce AutoHair , the first fully automatic method for 3D hair modeling from a single portrait image, with no user interaction or parameter tuning. Our method efficiently generates complete and high-quality hair geometries, which are comparable to those generated by the state-of-the-art methods, where user interaction is required. The core components of our method are: a novel hierarchical deep neural network for automatic hair segmentation and hair growth direction estimation, trained over an annotated hair image database; and an efficient and automatic data-driven hair matching and modeling algorithm, based on a large set of 3D hair exemplars. We demonstrate the efficacy and robustness of our method on Internet photos, resulting in a database of around 50K 3D hair models and a corresponding hairstyle space that covers a wide variety of real-world hairstyles. We also show novel applications enabled by our method, including 3D hairstyle space navigation and hair-aware image retrieval.

How to read this

Category
Method: fully automatic single-image 3D hair modeling
Contributions
  • Presents the first fully automatic single-image 3D hair modeling method, requiring no user interaction or parameter tuning
  • Introduces a hierarchical deep network for automatic hair segmentation and growth-direction estimation trained on an annotated hair-image database
  • Uses a data-driven matching and modeling algorithm over 3D hair exemplars, producing a ~50K-model database and a navigable hairstyle space, plus hair-aware image retrieval
Context
Builds on database-driven single-view hair modeling such as Hu et al. (hu-singleview-2015), removing the user interaction those methods required by adding learned segmentation and direction estimation.Builds on: Single-View Hair Modeling Using a Hairstyle Database
Correctness
Quality depends on the annotated training database and the coverage of the 3D hair-exemplar set; demonstrated on Internet photos with results stated as comparable to interactive state-of-the-art, so exotic or heavily occluded hairstyles outside the exemplar space are a likely limitation.
Clarity
Accessible; a first pass conveys the automatic pipeline, a second pass clarifies the network and matching algorithm.
How to read it
Focus first on how the deep segmentation/direction step feeds the exemplar matching; a second pass pays off for the network design and the hairstyle-space construction if hair modeling is your area.

Builds on

Built upon by

Related work

Keywords

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →