Learning Noise-Aware Encoder-Decoder from Noisy Labels by Alternating Back-Propagation for Saliency Detection

被引:65
作者
Zhang, Jing [1 ,3 ,4 ]
Xie, Jianwen [2 ]
Barnes, Nick [1 ]
机构
[1] Australian Natl Univ, Canberra, ACT, Australia
[2] Baidu Res, Cognit Comp Lab, Sunnyvale, CA USA
[3] Australian Ctr Robot Vis, Brisbane, Qld, Australia
[4] Data61, Eveleigh, Australia
来源
COMPUTER VISION - ECCV 2020, PT XVII | 2020年 / 12362卷
关键词
Noisy saliency; Latent variable model; Langevin dynamics; Alternating back-propagation; OBJECT DETECTION;
D O I
10.1007/978-3-030-58520-4_21
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In this paper, we propose a noise-aware encoder-decoder framework to disentangle a clean saliency predictor from noisy training examples, where the noisy labels are generated by unsupervised handcrafted feature-based methods. The proposed model consists of two sub-models parameterized by neural networks: (1) a saliency predictor that maps input images to clean saliency maps, and (2) a noise generator, which is a latent variable model that produces noises from Gaussian latent vectors. The whole model that represents noisy labels is a sum of the two sub-models. The goal of training the model is to estimate the parameters of both sub-models, and simultaneously infer the corresponding latent vector of each noisy label. We propose to train the model by using an alternating back-propagation (ABP) algorithm, which alternates the following two steps: (1) learning back-propagation for estimating the parameters of two sub-models by gradient ascent, and (2) inferential back-propagation for inferring the latent vectors of training noisy examples by Langevin Dynamics. To prevent the network from converging to trivial solutions, we utilize an edge-aware smoothness loss to regularize hidden saliency maps to have similar structures as their corresponding images. Experimental results on several benchmark datasets indicate the effectiveness of the proposed model.
引用
收藏
页码:349 / 366
页数:18
相关论文
共 57 条
[1]   Multiscale Combinatorial Grouping [J].
Arbelaez, Pablo ;
Pont-Tuset, Jordi ;
Barron, Jonathan T. ;
Marques, Ferran ;
Malik, Jitendra .
2014 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2014, :328-335
[2]  
Baidu, PaddlePaddle
[3]   Salient Object Detection: A Benchmark [J].
Borji, Ali ;
Cheng, Ming-Ming ;
Jiang, Huaizu ;
Li, Jia .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2015, 24 (12) :5706-5722
[4]   SalientShape: group saliency in image collections [J].
Cheng, Ming-Ming ;
Mitra, Niloy J. ;
Huang, Xiaolei ;
Hu, Shi-Min .
VISUAL COMPUTER, 2014, 30 (04) :443-453
[5]   Structure-measure: A New Way to Evaluate Foreground Maps [J].
Fan, Deng-Ping ;
Cheng, Ming-Ming ;
Liu, Yun ;
Li, Tao ;
Borji, Ali .
2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), 2017, :4558-4567
[6]  
Fan DP, 2018, PROCEEDINGS OF THE TWENTY-SEVENTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, P698
[7]   Attentive Feedback Network for Boundary-Aware Salient Object Detection [J].
Feng, Mengyang ;
Lu, Huchuan ;
Ding, Errui .
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019), 2019, :1623-1632
[8]  
Gold JR, 2017, PLAN HIST ENVIRON SE, P1
[9]  
Han T, 2017, AAAI CONF ARTIF INTE, P1976
[10]   Deep Residual Learning for Image Recognition [J].
He, Kaiming ;
Zhang, Xiangyu ;
Ren, Shaoqing ;
Sun, Jian .
2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2016, :770-778