← ArchivePaper2016
AutoHair: Fully Automatic Hair Modeling from a Single Image
Menglei Chai, Tianjia Shao, Hongzhi Wu, Yanlin Weng, Kun Zhou
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
Related work
- HairNet: Single-View Hair Reconstruction Using Convolutional Neural Networks 2018 / ECCV
- NeuralHDHair: Automatic High-Fidelity Hair Modeling from a Single Image Using Implicit Neural Representations 2022 / CVPR
- Single-View Hair Modeling Using a Hairstyle Database 2015 / SIGGRAPH
- Data-Driven Estimation of Cloth Simulation Models 2012 / CGF
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 →