Relighting Images in the Wild with a Self-Supervised Siamese Auto-Encoder

被引:8
|
作者
Liu, Yang [1 ]
Neophytou, Alexandros [2 ]
Sengupta, Sunando [2 ]
Sommerlade, Eric [2 ]
机构
[1] Univ Surrey, Guildford, Surrey, England
[2] Microsoft Corp, Reading, Berks, England
来源
2021 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV 2021) | 2021年
关键词
D O I
10.1109/WACV48630.2021.00008
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We propose a self-supervised method for image relighting of single view images in the wild. The method is based on an auto-encoder which deconstructs an image into two separate encodings, relating to the scene illumination and content, respectively. In order to disentangle this embedding information without supervision, we exploit the assumption that some augmentation operations do not affect the image content and only affect the direction of the light. A novel loss function, called spherical harmonic loss, is introduced that forces the illumination embedding to convert to a spherical harmonic vector. We train our model on large-scale datasets such as Youtube 8M and CelebA. Our experiments show that our method can correctly estimate scene illumination and realistically re-light input images, without any supervision or a prior shape model. Compared to supervised methods, our approach has similar performance and avoids common lighting artifacts.
引用
收藏
页码:32 / 40
页数:9
相关论文
共 50 条
  • [21] Stacked Fusion Supervised Auto-encoder with an Additional Classification Layer
    Rui Li
    Xiaodan Wang
    Wen Quan
    Lei Lei
    Neural Processing Letters, 2020, 51 : 2649 - 2667
  • [22] Disentangled Variational Auto-Encoder for semi-supervised learning
    Li, Yang
    Pan, Quan
    Wang, Suhang
    Peng, Haiyun
    Yang, Tao
    Cambria, Erik
    INFORMATION SCIENCES, 2019, 482 : 73 - 85
  • [23] Pairwise Supervised Hashing with Bernoulli Variational Auto-Encoder and Self-Control Gradient Estimator
    Dadaneh, Siamak Zamani
    Boluki, Shahin
    Yin, Mingzhang
    Zhou, Mingyuan
    Qian, Xiaoning
    CONFERENCE ON UNCERTAINTY IN ARTIFICIAL INTELLIGENCE (UAI 2020), 2020, 124 : 540 - 549
  • [24] Self-supervised transformer-based pre-training method using latent semantic masking auto-encoder for pest and disease classification
    Liu, Honglin
    Zhan, Yongzhao
    Xia, Huifen
    Mao, Qirong
    Tan, Yixin
    COMPUTERS AND ELECTRONICS IN AGRICULTURE, 2022, 203
  • [25] A road segmentation method based on the deep auto-encoder with supervised learning
    Song, Xiaona
    Rui, Ting
    Zhang, Sai
    Fei, Jianchao
    Wang, Xinqing
    COMPUTERS & ELECTRICAL ENGINEERING, 2018, 68 : 381 - 388
  • [26] Convolutional dynamic auto-encoder: a clustering method for semantic images
    Mohamed, Zahra
    Ksantini, Riadh
    Kaabi, Jihene
    NEURAL COMPUTING & APPLICATIONS, 2022, 34 (19): : 17087 - 17105
  • [27] Anomaly Detection for Medical Images Using Heterogeneous Auto-Encoder
    Lu, Shuai
    Zhang, Weihang
    Zhao, He
    Liu, Hanruo
    Wang, Ningli
    Li, Huiqi
    IEEE TRANSACTIONS ON IMAGE PROCESSING, 2024, 33 : 2770 - 2782
  • [28] Continual Representation Learning for Images with Variational Continual Auto-Encoder
    Jeon, Ik Hwan
    Shin, Soo Young
    PROCEEDINGS OF THE 11TH INTERNATIONAL CONFERENCE ON AGENTS AND ARTIFICIAL INTELLIGENCE (ICAART), VOL 2, 2019, : 367 - 373
  • [29] Convolutional dynamic auto-encoder: a clustering method for semantic images
    Zahra Mohamed
    Riadh Ksantini
    Jihene Kaabi
    Neural Computing and Applications, 2022, 34 : 17087 - 17105
  • [30] Convolutional dynamic auto-encoder: a clustering method for semantic images
    Mohamed, Zahra
    Ksantini, Riadh
    Kaabi, Jihene
    Neural Computing and Applications, 2022, 34 (19) : 17087 - 17105