3DBooSTeR: 3D Body Shape and Texture Recovery

Abstract

We propose 3DBooSTeR, a novel method to recover a textured 3D body mesh from a textured partial 3D scan. With the advent of virtual and augmented reality, there is a demand for creating realistic and high-fidelity digital 3D human representations. However, 3D scanning systems can only capture the 3D human body shape up to some level of defects due to its complexity, including occlusion between body parts, varying levels of details, shape deformations and the articulated skeleton. Textured 3D mesh completion is thus important to enhance 3D acquisitions. The proposed approach decouples the shape and texture completion into two sequential tasks. The shape is recovered by an encoder-decoder network deforming a template body mesh. The texture is subsequently obtained by projecting the partial texture onto the template mesh before inpainting the corresponding texture map with a novel approach. The approach is validated on the 3DBodyTex.v2 dataset.

SHARP 2020: The 1st Shape Recovery from Partial Textured 3D Scans Challenge Results

“…SnT-3DB comes second with 62%. RVH-IF surpasses the baseline unmodified partial data with an overall score of 46% higher and performs significantly better than SnT-3DB [14] with a score increment of 24%. RVH-IF-2 has similar shape and texture scores with differences of 2-3%, while SnT-3DB has a much higher texture score, 16% above the shape score.…”

“…Figures on validated registrations and entries for the challenges of SHARP 2020.The accepted entries are Implicit Feature Networks for Texture Completion of 3D Data[6,5], from RVH (Real Virtual Humans group at Max Planck Institute for Informatics), submitted in several variants to both Challenge 1 and Challenge 2, and 3DBooSTeR: 3D Body Shape and Texture Recovery[14], from SnT (Interdisciplinary Centre for Security, Reliability and Trust at the University of Luxembourg), submitted to Challenge 1. In the following, the entries are interchangeably abbreviated RVH-IF and SnT-3DB, respectively.…”

BCom-Net: Coarse-to-Fine 3D Textured Body Shape Completion Network

“…Aware of this need, recent competitions, such as SHApe Recovery from Partial Textured 3D Scans (SHARP) challenges [3,1,37], emerged in the research community to foster the development of joint shape and texture completion techniques from partial 3D acquisitions. The results of these competitions showed promising techniques [9,36] with a room for improvements.…”

“…Despite the impressive results in SHARP challenges [37,1], the method [9] does not output high-resolution texture and often over-smooths vertex colors over the whole shape. On the other hand, 3DBooSTeR method [36] decouples the problems of shape and texture completions and solves them in a sequential model. The shape completion of [36] is based on body models [12], while the texture completion is tackled as an image based texture-atlas inpainting.…”

“…On the other hand, 3DBooSTeR method [36] decouples the problems of shape and texture completions and solves them in a sequential model. The shape completion of [36] is based on body models [12], while the texture completion is tackled as an image based texture-atlas inpainting. [36] can complete high-resolution partial texture, but suffers from large shape and texture modelling artifacts.…”