Towards Unsupervised Learning of Generative Models for 3D Controllable Image Synthesis

被引:51
|
作者
Liao, Yiyi [1 ,2 ]
Schwarz, Katja [1 ,2 ]
Mescheder, Lars [1 ,2 ,3 ]
Geiger, Andreas [1 ,2 ]
机构
[1] Max Planck Inst Intelligent Syst, Tubingen, Germany
[2] Univ Tubingen, Tubingen, Germany
[3] Amazon, Tubingen, Germany
关键词
D O I
10.1109/CVPR42600.2020.00591
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In recent years, Generative Adversarial Networks have achieved impressive results in photorealistic image synthesis. This progress nurtures hopes that one day the classical rendering pipeline can be replaced by efficient models that are learned directly from images. However, current image synthesis models operate in the 2D domain where disentangling 3D properties such as camera viewpoint or object pose is challenging. Furthermore, they lack an interpretable and controllable representation. Our key hypothesis is that the image generation process should be modeled in 3D space as the physical world surrounding us is intrinsically three-dimensional. We define the new task of 3D controllable image synthesis and propose an approach for solving it by reasoning both in 3D space and in the 2D image domain. We demonstrate that our model is able to disentangle latent 3D factors of simple multi-object scenes in an unsupervised fashion from raw images. Compared to pure 2D baselines, it allows for synthesizing scenes that are consistent wrt. changes in viewpoint or object pose. We further evaluate various 3D representations in terms of their usefulness for this challenging task.
引用
收藏
页码:5870 / 5879
页数:10
相关论文
共 50 条
  • [1] Style Equalization: Unsupervised Learning of Controllable Generative Sequence Models
    Chang, Jen-Hao Rick
    Shrivastava, Ashish
    Koppula, Hema Swetha
    Zhang, Xiaoshuai
    Tuzel, Oncel
    [J]. INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 162, 2022,
  • [2] Learning Generative Models of 3D Structures
    Chaudhuri, Siddhartha
    Ritchie, Daniel
    Wu, Jiajun
    Xu, Kai
    Zhang, Hao
    [J]. COMPUTER GRAPHICS FORUM, 2020, 39 (02) : 643 - 666
  • [3] VeRi3D: Generative Vertex-based Radiance Fields for 3D Controllable Human Image Synthesis
    Chen, Xinya
    Huang, Jiaxin
    Bin, Yanrui
    Yu, Lu
    Liao, Yiyi
    [J]. 2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, : 8952 - 8963
  • [4] SFGAN: Unsupervised Generative Adversarial Learning of 3D Scene Flow from the 3D Scene Self
    Wang, Guangming
    Jiang, Chaokang
    Shen, Zehang
    Miao, Yanzi
    Wang, Hesheng
    [J]. ADVANCED INTELLIGENT SYSTEMS, 2022, 4 (04)
  • [5] Learning Representations and Generative Models for 3D Point Clouds
    Achlioptas, Panos
    Diamanti, Olga
    Mitliagkas, Ioannis
    Guibas, Leonidas
    [J]. INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 80, 2018, 80
  • [6] 3D MR image synthesis using unsupervised deep learning algorithm in MS patients
    Jaberzadeh, Amir
    Rukmangadachar, Lokesh
    Pelletier, Daniel
    [J]. NEUROLOGY, 2019, 92 (15)
  • [7] Generative VoxelNet: Learning Energy-Based Models for 3D Shape Synthesis and Analysis
    Xie, Jianwen
    Zheng, Zilong
    Gao, Ruiqi
    Wang, Wenguan
    Zhu, Song-Chun
    Wu, Ying Nian
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2022, 44 (05) : 2468 - 2484
  • [8] Unsupervised Segmentation of 3D Microvascular Photoacoustic Images Using Deep Generative Learning
    Sweeney, Paul W.
    Hacker, Lina
    Lefebvre, Thierry L.
    Brown, Emma L.
    Grohl, Janek
    Bohndiek, Sarah E.
    [J]. ADVANCED SCIENCE, 2024, 11 (32)
  • [9] Unsupervised Learning with Alternating Matching Features for 3D Image Registration
    Liu, Lin
    Fan, Xinxin
    Liu, Xuan
    He, Wenfeng
    Liang, Xiaokun
    [J]. MEDICAL IMAGING 2024: IMAGE PROCESSING, 2024, 12926
  • [10] SemanticStyleGAN: Learning Compositional Generative Priors for Controllable Image Synthesis and Editing
    Shi, Yichun
    Yang, Xiao
    Wan, Yangyue
    Shen, Xiaohui
    [J]. 2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2022, : 11244 - 11254