Filters








1,774 Hits in 9.3 sec

Learning by Analogy: Reliable Supervision from Transformations for Unsupervised Optical Flow Estimation [article]

Liang Liu, Jiangning Zhang, Ruifei He, Yong Liu, Yabiao Wang, Ying Tai, Donghao Luo, Chengjie Wang, Jilin Li, Feiyue Huang
2020 arXiv   pre-print
Unsupervised learning of optical flow, which leverages the supervision from view synthesis, has emerged as a promising alternative to supervised methods.  ...  However, the objective of unsupervised learning is likely to be unreliable in challenging scenes. In this work, we present a framework to use more reliable supervision from transformations.  ...  Acknowledgment We thank anonymous reviewers for their constructive comments, and LL would like to thank Pengpeng Liu for helpful suggestions.  ... 
arXiv:2003.13045v2 fatcat:obz5icpdjvbktmaq36633p3ple

Learning by Analogy: Reliable Supervision From Transformations for Unsupervised Optical Flow Estimation

Liang Liu, Jiangning Zhang, Ruifei He, Yong Liu, Yabiao Wang, Ying Tai, Donghao Luo, Chengjie Wang, Jilin Li, Feiyue Huang
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Unsupervised learning of optical flow, which leverages the supervision from view synthesis, has emerged as a promising alternative to supervised methods.  ...  However, the objective of unsupervised learning is likely to be unreliable in challenging scenes. In this work, we present a framework to use more reliable supervision from transformations.  ...  Acknowledgment We thank anonymous reviewers for their constructive comments, and LL would like to thank Pengpeng Liu for helpful suggestions.  ... 
doi:10.1109/cvpr42600.2020.00652 dblp:conf/cvpr/LiuZHLWTLWLH20 fatcat:fbx2rhydvnfgpd7r6fg7bq5wfu

Unsupervised Learning of Landmarks by Descriptor Vector Exchange

James Thewlis, Samuel Albanie, Hakan Bilen, Andrea Vedaldi
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
Abstract Equivariance to random image transformations is an effective method to learn landmarks of object categories, such as the eyes and the nose in faces, without manual supervision.  ...  The embeddings learned for the category of faces are visualised in the figure above with the help of a query image [8], shown in the centre of the figure.  ...  We thank Almut Sophia Koepke for helpful discussions. We are grateful to ERC StG IDIU-638009, EP/R03298X/1 and AWS Machine Learning Research Awards (MLRA) for support.  ... 
doi:10.1109/iccv.2019.00646 dblp:conf/iccv/ThewlisABV19 fatcat:sumqz5oudjep5kmqcxuistehqe

Discovery of Visual Semantics by Unsupervised and Self-Supervised Representation Learning [article]

Gustav Larsson
2017 arXiv   pre-print
We show that traditional methods for unsupervised learning, such as layer-wise clustering or autoencoders, remain inferior to supervised pre-training.  ...  The success of deep learning in computer vision is rooted in the ability of deep networks to scale up model complexity as demanded by challenging visual tasks.  ...  [90] learn general-purpose representation by predicting saliency based on optical flow. Owens et al.  ... 
arXiv:1708.05812v1 fatcat:w77w3q3ms5c5fnyzl65mkj4ozy

Multivariate analysis of Brillouin imaging data by supervised and unsupervised learning [article]

YuChen Xiang, Kai Ling C. Seow, Carl Paterson, Peter Török
2020 arXiv   pre-print
The estimated spectral parameters are consistent with those calculated from pure fitting.  ...  Brillouin imaging relies on the reliable extraction of subtle spectral information from hyperspectral datasets.  ...  In machine learning, algorithms are commonly classified as supervised or unsupervised.  ... 
arXiv:2009.07220v1 fatcat:5pce5i2uyzfenpwuvpn4v4sbl4

Video Abnormal Event Detection by Learning to Complete Visual Cloze Tests [article]

Siqi Wang, Guang Yu, Zhiping Cai, Xinwang Liu, En Zhu, Jianping Yin
2021 arXiv   pre-print
Then, the VCT is completed by training DNNs to infer the erased patch and its optical flow via video semantics.  ...  For each marked region, a normalized patch sequence is extracted from current and adjacent frames and stacked into a STC.  ...  Optical flow can be computed by either classic methods or DNN based methods [16] . For efficiency, we estimate the dense optical flow by a pre-trained FlowNetv2 model [20] .  ... 
arXiv:2108.02356v2 fatcat:7sl2musf7vecrdtdhwcqb2nsjy

UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss [article]

Simon Meister, Junhwa Hur, Stefan Roth
2017 arXiv   pre-print
Inspired by classical energy-based optical flow methods, we design an unsupervised loss based on occlusion-aware bidirectional flow estimation and the robust census transform to circumvent the need for  ...  Therefore, recent end-to-end convolutional networks for optical flow rely on synthetic datasets for supervision, but the domain mismatch between training and test scenarios continues to be a challenge.  ...  Acknowledgements The research leading to these results has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013) / ERC Grant Agreement  ... 
arXiv:1711.07837v1 fatcat:i5lmezoaivdfdfbmvzptbwhedy

UnFlow: Unsupervised Learning of Optical Flow With a Bidirectional Census Loss

Simon Meister, Junhwa Hur, Stefan Roth
2018 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
Inspired by classical energy-based optical flow methods, we design an unsupervised loss based on occlusion-aware bidirectional flow estimation and the robust census transform to circumvent the need for  ...  Therefore, recent end-to-end convolutional networks for optical flow rely on synthetic datasets for supervision, but the domain mismatch between training and test scenarios continues to be a challenge.  ...  Acknowledgements The research leading to these results has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013) / ERC Grant Agreement  ... 
doi:10.1609/aaai.v32i1.12276 fatcat:y5zmvtexpngktcv2rhz4ig3s3i

Warp Consistency for Unsupervised Learning of Dense Correspondences [article]

Prune Truong and Martin Danelljan and Fisher Yu and Luc Van Gool
2021 arXiv   pre-print
We propose Warp Consistency, an unsupervised learning objective for dense correspondence regression. Our objective is effective even in settings with large appearance and view-point changes.  ...  The key challenge in learning dense correspondences lies in the lack of ground-truth matches for real image pairs.  ...  Acknowledgements: This work was supported by the ETH Zürich Fund (OK), a Huawei Gift, the ETH Future Computing Laboratory (EFCL) financed by a gift from Huawei Technologies, Amazon AWS, and an Nvidia GPU  ... 
arXiv:2104.03308v3 fatcat:wgukjdkv3ngydgvdsstt6pq7om

Vision Goes Symbolic Without Loss of Information Within the Preattentive Vision Phase: The Need to Shift the Learning Paradigm from Machine-Learning (from Examples) to Machine-Teaching (by Rules) at the First Stage of a Two-Stage Hybrid Remote... Part I [chapter]

Andrea Baraldi
2012 Earth Observation  
Statistical models are capable of learning from either labeled (supervised) or unlabeled (unsupervised) data, refer to Fig. 1 .  ...  Unlabeled (unsupervised) data learning Labeled (supervised) data learning Data quantization Entropy maximization Probability density function estimation Classification Function regression  ...  Vision Goes Symbolic Without Loss of Information Within the Preattentive Vision Phase: The Need to Shift the Learning Paradigm from Machine-Learning (from Examples) to Machine-Teaching (by Rules) at the  ... 
doi:10.5772/34035 fatcat:babno7wq3vfpthjvswxuvpodlm

Digging into Uncertainty in Self-supervised Multi-view Stereo [article]

Hongbin Xu, Zhipeng Zhou, Yali Wang, Wenxiong Kang, Baigui Sun, Hao Li, Yu Qiao
2021 arXiv   pre-print
To this end, we propose to estimate epistemic uncertainty in self-supervised MVS, accounting for what the model ignores.  ...  To address these issues, we propose a novel Uncertainty reduction Multi-view Stereo (UMVS) framework for self-supervised learning.  ...  The two-view pairs for optical flow estimation are selected by combining the reference view with each of the source views provided by MVS-Net [39] .  ... 
arXiv:2108.12966v2 fatcat:bzcpxvbzazh37n37ce6o4hokne

Unsupervised Deep Homography: A Fast and Robust Homography Estimation Model [article]

Ty Nguyen, Steven W. Chen, Shreyas S. Shivakumar, Camillo J. Taylor, Vijay Kumar
2018 arXiv   pre-print
In this study, we propose an unsupervised learning algorithm that trains a Deep Convolutional Neural Network to estimate planar homographies.  ...  Homography estimation between multiple aerial images can provide relative pose estimation for collaborative autonomous exploration and monitoring.  ...  ACKNOWLEDGEMENTS We gratefully acknowledge the support of ARL grants W911NF-08-2-0004 and W911NF-10 We would also like to thank Aerial Applications for the UAV data set.  ... 
arXiv:1709.03966v3 fatcat:ndaw3eapdfedlk2tfnvkf2bbte

Approaches, Challenges, and Applications for Deep Visual Odometry: Toward to Complicated and Emerging Areas [article]

Ke Wang, Sai Ma, Junlan Chen, Fan Ren
2020 arXiv   pre-print
Comparing with classical geometry-based methods, deep learning-based methods can automatically learn effective and robust representations, such as depth, optical flow, feature, ego-motion, etc., from data  ...  Then, using the offered criteria as the uniform measurements, we detailedly evaluate and discuss how deep learning improves the performance of VO from the aspects of depth estimation, feature extraction  ...  DeMoN [96] was the first to use unsupervised learning methods to estimate both depth and pose from consecutive images. This network used optical flow to assist the depth and motion estimation.  ... 
arXiv:2009.02672v1 fatcat:zdnwt4lpmvbiromtpcxhxxpjxa

Self-Supervised Pillar Motion Learning for Autonomous Driving [article]

Chenxu Luo, Xiaodong Yang, Alan Yuille
2021 arXiv   pre-print
To this end, we propose a learning framework that leverages free supervisory signals from point clouds and paired camera images to estimate motion purely via self-supervision.  ...  In this paper, we seek to answer the research question of whether the abundant unlabeled data collections can be utilized for accurate and efficient motion learning.  ...  to reliably recover the static points from optical flow.  ... 
arXiv:2104.08683v1 fatcat:g7xx35yltzafrobimjwnwxvcvm

Unsupervised Deep Video Denoising [article]

Dev Yashpal Sheth, Sreyas Mohan, Joshua L. Vincent, Ramon Manzorro, Peter A. Crozier, Mitesh M. Khapra, Eero P. Simoncelli, Carlos Fernandez-Granda
2021 arXiv   pre-print
Thus, the network learns to perform implicit motion compensation, even though it is only trained for denoising.  ...  Deep convolutional neural networks (CNNs) for video denoising are typically trained with supervision, assuming the availability of clean videos.  ...  This work was supported by HHMI, NSF NRT HDR Award 1922658, CBET 1604971, OAC-1940263 and OAC-1940097. We thank the HPC staff at NYU, ASU and RBCDSAI, IIT Madras for their support.  ... 
arXiv:2011.15045v3 fatcat:gg3lq7e2ebcfjcebjql3ifm5we
« Previous Showing results 1 — 15 out of 1,774 results