Style Agnostic 3D Reconstruction via Adversarial Style Transfer

被引:0
|
作者
Petersen, Felix [1 ]
Goldluecke, Bastian [1 ]
Deussen, Oliver [1 ]
Kuehne, Hilde [2 ,3 ]
机构
[1] Univ Konstanz, Constance, Germany
[2] Goethe Univ Frankfurt, Frankfurt, Germany
[3] IBM MIT Watson AI Lab, Cambridge, MA USA
关键词
D O I
10.1109/WACV51458.2022.00233
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Reconstructing the 3D geometry of an object from an image is a major challenge in computer vision. Recently introduced differentiable renderers can be leveraged to learn the 3D geometry of objects from 2D images, but those approaches require additional supervision to enable the renderer to produce an output that can be compared to the input image. This can be scene information or constraints such as object silhouettes, uniform backgrounds, material, texture, and lighting. In this paper, we propose an approach that enables a differentiable rendering-based learning of 3D objects from images with backgrounds without the need for silhouette supervision. Instead of trying to render an image close to the input, we propose an adversarial style-transfer and domain adaptation pipeline that allows to translate the input image domain to the rendered image domain. This allows us to directly compare between a translated image and the differentiable rendering of a 3D object reconstruction in order to train the 3D object reconstruction network. We show that the approach learns 3D geometry from images with backgrounds and provides a better performance than constrained methods for single-view 3D object reconstruction on this task.
引用
收藏
页码:2273 / 2282
页数:10
相关论文
共 50 条
  • [1] 3D Reconstruction Based on Style Transfer Data Augmentation
    Saruwatari T.
    Inoue K.
    Yoshioka M.
    IEEJ Transactions on Electronics, Information and Systems, 2020, 140 (11) : 1198 - 1206
  • [2] 3D Material Style Transfer
    Nguyen, Chuong H.
    Ritschel, Tobias
    Myszkowski, Karol
    Eisemann, Elmar
    Seidel, Hans-Peter
    COMPUTER GRAPHICS FORUM, 2012, 31 (02) : 431 - 438
  • [3] 3D image augmentation using Neural Style Transfer and Generative Adversarial Networks
    Cuong Do
    APPLICATIONS OF DIGITAL IMAGE PROCESSING XLIII, 2020, 11510
  • [4] 3D shape creation by style transfer
    Han, Zhizhong
    Liu, Zhenbao
    Han, Junwei
    Bu, Shuhui
    VISUAL COMPUTER, 2015, 31 (09): : 1147 - 1161
  • [5] 3D shape creation by style transfer
    Zhizhong Han
    Zhenbao Liu
    Junwei Han
    Shuhui Bu
    The Visual Computer, 2015, 31 : 1147 - 1161
  • [6] Neural style transfer for 3D meshes
    Kang, Hongyuan
    Dong, Xiao
    Cao, Juan
    Chen, Zhonggui
    GRAPHICAL MODELS, 2023, 129
  • [7] Selective Style Gaussian: Enabling Localized Style Transfer in 3D Scenes
    Yang, Hyeri
    Heo, Hyeonbeom
    Hong, Junyoung
    Kim, Ye Ju
    Lee, Kyungjae
    2024 INTERNATIONAL TECHNICAL CONFERENCE ON CIRCUITS/SYSTEMS, COMPUTERS, AND COMMUNICATIONS, ITC-CSCC 2024, 2024,
  • [8] Analogy-driven 3D style transfer
    Ma, Chongyang
    Huang, Haibin
    Sheffer, Alla
    Kalogerakis, Evangelos
    Wang, Rui
    COMPUTER GRAPHICS FORUM, 2014, 33 (02) : 175 - 184
  • [9] Indoor camera pose estimation via style-transfer 3D models
    Chen, Junjie
    Li, Shuai
    Liu, Donghai
    Lu, Weisheng
    COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING, 2022, 37 (03) : 335 - 353
  • [10] AdvST: Generating Unrestricted Adversarial Images via Style Transfer
    Wang, Xiaomeng
    Chen, Honglong
    Sun, Peng
    Li, Junjian
    Zhang, Anqing
    Liu, Weifeng
    Jiang, Nan
    IEEE TRANSACTIONS ON MULTIMEDIA, 2024, 26 : 4846 - 4858