Disentangled Human Body Embedding Based on Deep Hierarchical Neural Network
arXiv:1905.05622 · doi:10.1109/TVCG.2020.2988476
Abstract
Human bodies exhibit various shapes for different identities or poses, but the body shape has certain similarities in structure and thus can be embedded in a low-dimensional space. This paper presents an autoencoder-like network architecture to learn disentangled shape and pose embedding specifically for the 3D human body. This is inspired by recent progress of deformation-based latent representation learning. To improve the reconstruction accuracy, we propose a hierarchical reconstruction pipeline for the disentangling process and construct a large dataset of human body models with consistent connectivity for the learning of the neural network. Our learned embedding can not only achieve superior reconstruction accuracy but also provide great flexibility in 3D human body generation via interpolation, bilinear interpolation, and latent space sampling. The results from extensive experiments demonstrate the powerfulness of our learned 3D human body embedding in various applications.
This manuscript is accepted for publication in the IEEE Transactions on Visualization and Computer Graphics Journal (IEEE TVCG). The Code is available at https://github.com/Juyong/DHNN_BodyRepresentation
References in corpus (7)
- Adam: A Method for Stochastic Optimization
- Auto-Encoding Variational Bayes
- Embodied Hands: Modeling and Capturing Hands and Bodies Together
- Building Statistical Shape Spaces for 3D Human Modeling
- Multi-chart Generative Surface Modeling
- MonoPerfCap: Human Performance Capture from Monocular Video
- Sparse Data Driven Mesh Deformation
Cited by in corpus (8)
- Driving-Signal Aware Full-Body Avatars
- Multiway Non-rigid Point Cloud Registration via Learned Functional Map Synchronization
- Neural-ABC: Neural Parametric Models for Articulated Body with Clothes
- Neural Human Deformation Transfer
- BCNet: Learning Body and Cloth Shape from A Single Image
- StereoPIFu: Depth Aware Clothed Human Digitization via Stereo Vision
- Unsupervised Shape and Pose Disentanglement for 3D Meshes
- LatentHuman: Shape-and-Pose Disentangled Latent Representation for Human Bodies