Filters








75,623 Hits in 4.8 sec

Self-supervised Learning from a Multi-view Perspective [article]

Yao-Hung Hubert Tsai, Yue Wu, Ruslan Salakhutdinov, Louis-Philippe Morency
2021 arXiv   pre-print
Many proposed approaches for self-supervised learning follow naturally a multi-view perspective, where the input (e.g., original images) and the self-supervised signals (e.g., augmented images) can be  ...  Building from this multi-view perspective, this paper provides an information-theoretical framework to better understand the properties that encourage successful self-supervised learning.  ...  Our work shares a similar goal to explain the success of SSL, from the perspectives of Information Theory (Cover & Thomas, 2012) and multi-view representation 1 .  ... 
arXiv:2006.05576v4 fatcat:slhsl2b7lbcepnw56ehw7oc3c4

A Multi-view Perspective of Self-supervised Learning [article]

Chuanxing Geng, Zhenghao Tan, Songcan Chen
2020 arXiv   pre-print
As a newly emerging unsupervised learning paradigm, self-supervised learning (SSL) recently gained widespread attention, which usually introduces a pretext task without manual annotation of data.  ...  In this paper, we borrow a multi-view perspective to decouple a class of popular pretext tasks into a combination of view data augmentation (VDA) and view label classification (VLC), where we attempt to  ...  From a multi-view perspective, we can find that existing pretext tasks are mainly to perform explicit linear or nonlinear transformation (i.e., self-supervised signals) on the given original (single view  ... 
arXiv:2003.00877v2 fatcat:6bcsj52rdng3denuwjx33sfmz4

Time-Contrastive Networks: Self-Supervised Learning from Multi-view Observation

Pierre Sermanet, Corey Lynch, Jasmine Hsu, Sergey Levine
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
Both models are initialized with ImageNet classification weights, then trained in a self-supervised manner using 15 minutes of multi-view pouring videos (no labels).  ...  We also show what is, to the best of our knowledge, the first self-supervised results for end-to-end imitation learning of human motions with a real robot (Table 2) .  ...  The model is able to learn a complex human to robot mapping entirely self-supervised and is quantitatively better than a human-supervised imitation ( Table 2) .  ... 
doi:10.1109/cvprw.2017.69 dblp:conf/cvpr/SermanetLHL17 fatcat:zd7cicgfsjhzlig3rc2npeo4nq

Intra-Camera Supervised Person Re-Identification: A New Benchmark [article]

Xiangping Zhu and Xiatian Zhu and Minxian Li and Vittorio Murino and Shaogang Gong
2019 arXiv   pre-print
Under this ICS setting with weaker label supervision, we formulate a Multi-Task Multi-Label (MTML) deep learning method.  ...  This is achieved by inter-camera multi-label learning under a joint multi-task inference framework.  ...  To benefit model training from self-discovered identity matching pairs, a proper supervision function is designed.  ... 
arXiv:1908.10344v1 fatcat:fcupstaflvcgnitwr24lwipc7u

Self-supervised Learning of Depth Inference for Multi-view Stereo [article]

Jiayu Yang, Jose M. Alvarez, Miaomiao Liu
2021 arXiv   pre-print
Here, we propose a self-supervised learning framework for multi-view stereo that exploit pseudo labels from the input data.  ...  Extensive experiments on the DTU dataset show that our proposed self-supervised learning framework outperforms existing unsupervised multi-view stereo networks by a large margin and performs on par compared  ...  In this paper, we propose a self-supervised learning framework for depth inference from multi-view images.  ... 
arXiv:2104.02972v1 fatcat:cqcqznijafh7df5zjfokg35y3y

CR-GAN: Learning Complete Representations for Multi-view Generation

Yu Tian, Xi Peng, Long Zhao, Shaoting Zhang, Dimitris N. Metaxas
2018 Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence  
Generating multi-view images from a single-view input is an important yet challenging problem. It has broad applications in vision, graphics, and robotics.  ...  More importantly, the two-pathway framework makes it possible to combine both labeled and unlabeled data for self-supervised learning, which further enriches the embedding space for realistic generations  ...  , using a two-pathway learning scheme; • CR-GAN can leverage unlabeled data for self-supervised learning, yielding improved generation quality; • CR-GAN can generate high-quality multi-view images from  ... 
doi:10.24963/ijcai.2018/131 dblp:conf/ijcai/TianPZZM18 fatcat:rrmzcjbbyfdbjbvobqrmibttm4

CR-GAN: Learning Complete Representations for Multi-view Generation [article]

Yu Tian, Dimitris N. Metaxas University of North Carolina at Charlotte)
2018 arXiv   pre-print
Generating multi-view images from a single-view input is an essential yet challenging problem. It has broad applications in vision, graphics, and robotics.  ...  More importantly, the two-pathway framework makes it possible to combine both labeled and unlabeled data for self-supervised learning, which further enriches the embedding space for realistic generations  ...  , using a two-pathway learning scheme; • CR-GAN can leverage unlabeled data for self-supervised learning, yielding improved generation quality; • CR-GAN can generate high-quality multi-view images from  ... 
arXiv:1806.11191v1 fatcat:bro7smpffnbbvh5ldhjmfu6xcu

Self-supervised Discriminative Feature Learning for Deep Multi-view Clustering [article]

Jie Xu, Yazhou Ren, Huayi Tang, Zhimeng Yang, Lili Pan, Yang Yang, Xiaorong Pu
2021 arXiv   pre-print
In a self-supervised manner, pseudo-labels are obtained to build a unified target distribution to perform multi-view discriminative feature learning.  ...  To address this drawback, we propose self-supervised discriminative feature learning for deep multi-view clustering (SDMVC).  ...  Self-supervised learning is the recent hot topic of the community. The framework proposed in [38] combined a self-supervised paradigm with multi-view clustering.  ... 
arXiv:2103.15069v2 fatcat:5illq2c2qfa7jfvq76dnycxphq

Information Theory-Guided Heuristic Progressive Multi-View Coding [article]

Jiangmeng Li, Wenwen Qiang, Hang Gao, Bing Su, Farid Razzak, Jie Hu, Changwen Zheng, Hui Xiong
2021 arXiv   pre-print
To this end, we rethink the existing multi-view learning paradigm from the information theoretical perspective and then propose a novel information theoretical framework for generalized multi-view learning  ...  Multi-view representation learning captures comprehensive information from multiple views of a shared context.  ...  Conclusion We rethink the self-supervised MVL from the information theoretical perspective and then propose the information theoretical framework of generalized multi-view self-supervision.  ... 
arXiv:2109.02344v1 fatcat:ol4u6d7yafez5fpof4l5xb3nam

Multi-Scale Contrastive Siamese Networks for Self-Supervised Graph Representation Learning [article]

Ming Jin, Yizhen Zheng, Yuan-Fang Li, Chen Gong, Chuan Zhou, Shirui Pan
2021 arXiv   pre-print
Specifically, we first generate two augmented views from the input graph based on local and global perspectives.  ...  to learn node representations by enhancing Siamese self-distillation with multi-scale contrastive learning.  ...  Graph Augmentations Augmentation is a key component in self-supervised visual representation learning.  ... 
arXiv:2105.05682v2 fatcat:57t7mylspbbs3hb4wf3h5jsgoy

Unsupervised 3D Pose Estimation With Geometric Self-Supervision

Ching-Hang Chen, Ambrish Tyagi, Amit Agrawal, Dylan Drover, Rohith MV, Stefan Stojanov, James M. Rehg
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
We present an unsupervised learning approach to recover 3D human pose from 2D skeletal joints extracted from a single image.  ...  The training can thus be self supervised by exploiting the geometric selfconsistency of the lift-reproject-lift process.  ...  [36] proposed an unsupervised method to learn a geometry-aware body representation. Their approach maps one view of the human to another view from a set of given multi-view images.  ... 
doi:10.1109/cvpr.2019.00586 dblp:conf/cvpr/ChenTADMSR19 fatcat:rdnsec26urgqbmekearsud3t4y

Socially Supervised Representation Learning: the Role of Subjectivity in Learning Efficient Representations [article]

Julius Taylor, Eleni Nisioti, Clément Moulin-Frier
2021 arXiv   pre-print
Despite its rise as a prominent solution to the data inefficiency of today's machine learning models, self-supervised learning has yet to be studied from a purely multi-agent perspective.  ...  Altogether, our results demonstrate how communication from subjective perspectives can lead to the acquisition of more abstract representations in multi-agent systems, opening promising perspectives for  ...  This work also benefited from access to the HPC resources of IDRIS under the allocation 2020-[A0091011996] made by GENCI, using the Jean Zay supercomputer.  ... 
arXiv:2109.09390v3 fatcat:46j2mvmxivglnh6di3uc5w6ywa

Self-supervised Feature Learning by Cross-modality and Cross-view Correspondences [article]

Longlong Jing, Yucheng Chen, Ling Zhang, Mingyi He, Yingli Tian
2020 arXiv   pre-print
Unlike most existing self-supervised methods to learn only 2D image features or only 3D point cloud features, this paper presents a novel and effective self-supervised learning approach to jointly learn  ...  The effectiveness of the learned 2D and 3D features is evaluated by transferring them on five different tasks including multi-view 2D shape recognition, 3D shape recognition, multi-view 2D shape retrieval  ...  p from self-supervised pretext task can be served as a good starting point for the optimization.  ... 
arXiv:2004.05749v1 fatcat:fbpilwf3hjaxxdpulizzwjmnfa

Geometry-Driven Self-Supervised Method for 3D Human Pose Estimation

Yang Li, Kan Li, Shuai Jiang, Ziyue Zhang, Congzhentao Huang, Richard Yi Da Xu
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
In this paper, we propose a novel self-supervised approach to avoid the need of manual annotations.  ...  Besides, we adopt the confidences of 2D joints to integrate losses from different views to alleviate the influence of noises caused by the self-occlusion problem.  ...  pre-trained an encoderdecoder network that predicts an image from one view to another to learn a geometry-aware body representation, and then use a small amount of supervision to learn a mapping from 2D  ... 
doi:10.1609/aaai.v34i07.6808 fatcat:gge77eow2nhl3lfoqgqc33nyku

Intra-Camera Supervised Person Re-Identification [article]

Xiangping Zhu, Xiatian Zhu, Minxian Li, Pietro Morerio, Vittorio Murino, Shaogang Gong
2021 arXiv   pre-print
Consequently, it gives rise to a more scalable and more feasible setting, which we call Intra-Camera Supervised (ICS) person re-id, for which we formulate a Multi-tAsk mulTi-labEl (MATE) deep learning  ...  supervised learning competitors.  ...  The key idea is to self-mine supervision information from unlabelled training data based on the knowledge learned from a small proportion of labelled training data.  ... 
arXiv:2002.05046v3 fatcat:g6ikgbqvxzgsnp5qzg6rmqrfqq
« Previous Showing results 1 — 15 out of 75,623 results