Filters








7,073 Hits in 1.5 sec

Unsupervised Domain Adaptation for Visual Navigation [article]

Shangda Li, Devendra Singh Chaplot, Yao-Hung Hubert Tsai, Yue Wu, Louis-Philippe Morency, Ruslan Salakhutdinov
2020 arXiv   pre-print
In order for these policies to be practically useful, they need to be transferred to the real-world. In this paper, we propose an unsupervised domain adaptation method for visual navigation.  ...  Our method translates the images in the target domain to the source domain such that the translation is consistent with the representations learned by the navigation policy.  ...  Among domain adaptation techniques, unsupervised methods are favorable because it is extremely expensive to collect parallel data for the purpose of visual navigation.  ... 
arXiv:2010.14543v2 fatcat:evtvmq4wg5gyzoldvwwaupjxde

When Autonomous Systems Meet Accuracy and Transferability through AI: A Survey

Chongzhen Zhang, Jianrui Wang, Gary G. Yen, Chaoqiang Zhao, Qiyu Sun, Yang Tang, Feng Qian, Jürgen Kurths
2020 Patterns  
Transferability means that when a well-trained model is transferred to other testing domains, the accuracy is still good.  ...  Finally, we discuss several challenges and future topics for the use of adversarial learning, RL, and meta-learning in autonomous systems.  ...  ACKNOWLEDGMENTS The authors would like to thank the Editor-in-Chief, Scientific Editor, and anonymous referees for their helpful comments and suggestions, which have greatly improved this paper.  ... 
doi:10.1016/j.patter.2020.100050 pmid:33205114 pmcid:PMC7660378 fatcat:vs7wm2yrwjamjbaml36663wvze

When Autonomous Systems Meet Accuracy and Transferability through AI: A Survey [article]

Chongzhen Zhang, Jianrui Wang, Gary G. Yen, Chaoqiang Zhao, Qiyu Sun, Yang Tang, Feng Qian, Jürgen Kurths
2020 arXiv   pre-print
Transferability means that when a well-trained model is transferred to other testing domains, the accuracy is still good.  ...  Finally, we discuss several challenges and future topics for using adversarial learning, RL and meta-learning in autonomous systems.  ...  Domain adaptation for visual applications includes shallow and deep methods [31] .  ... 
arXiv:2003.12948v3 fatcat:qtmjs74p2vh6thdotbhgebdvoi

On Embodied Visual Navigation in Real Environments Through Habitat [article]

Marco Rosano, Antonino Furnari, Luigi Gulino, Giovanni Maria Farinella
2020 arXiv   pre-print
We perform a range of experiments to assess the ability of such policies to generalize using virtual and real-world images, as well as observations transformed with unsupervised domain adaptation approaches  ...  Visual navigation models based on deep learning can learn effective policies when trained on large amounts of visual observations through reinforcement learning.  ...  We found that both unsupervised domain adaptation such as CycleGAN [11] and supervised adaptation such as fine-tuning, allow to adapt the navigation policy to the real-world scenario, with the latter  ... 
arXiv:2010.13439v1 fatcat:jbxdrogrureh5jquhqjjlahpsu

BoMuDANet: Unsupervised Adaptation for Visual Scene Understanding in Unstructured Driving Environments [article]

Divya Kothandaraman, Rohan Chandra, Dinesh Manocha
2021 arXiv   pre-print
We present an unsupervised adaptation approach for visual scene understanding in unstructured traffic environments.  ...  We describe a new semantic segmentation technique based on unsupervised domain adaptation (DA), that can identify the class or category of each region in RGB images or videos.  ...  [4] (Comb.) refers to the combined approach for boundless unsupervised domain adaptation ("BUDA").  ... 
arXiv:2010.03523v3 fatcat:72sr7avhhjb7niwkheizlc2eeu

Unsupervised Curricula for Visual Meta-Reinforcement Learning [article]

Allan Jabri, Kyle Hsu, Ben Eysenbach, Abhishek Gupta, Sergey Levine, Chelsea Finn
2019 arXiv   pre-print
We develop an unsupervised algorithm for inducing an adaptive meta-training task distribution, i.e. an automatic curriculum, by modeling unsupervised interaction in a visual environment.  ...  In experiments on vision-based navigation and manipulation domains, we show that the algorithm allows for unsupervised meta-learning that transfers to downstream tasks specified by hand-crafted reward  ...  visual navigation.  ... 
arXiv:1912.04226v1 fatcat:cer4sn5jm5ezrfmcvsvfs3t6ii

Unsupervised Domain Adaptation Learning Algorithm for RGB-D Staircase Recognition [article]

Jing Wang, Kuangen Zhang
2019 arXiv   pre-print
In this paper, we apply an unsupervised domain adaptation approach in deep architectures to transfer knowledge learned from the labeled RGB-D stationary staircase dataset to the unlabeled RGB-D escalator  ...  We demonstrate the success of the approach for classifying staircase on two domains with a limited amount of data.  ...  Unsupervised Domain Adaptation Learning For classification tasks, we have the input space X and the set of L possible labels Y.  ... 
arXiv:1903.01212v4 fatcat:5clqxnkmajh5fehvudqrxg4zea

VisDA-2021 Competition Universal Domain Adaptation to Improve Performance on Out-of-Distribution Data [article]

Dina Bashkirova, Dan Hendrycks, Donghyun Kim, Samarth Mishra, Kate Saenko, Kuniaki Saito, Piotr Teterwak, Ben Usman
2021 arXiv   pre-print
We set up unsupervised domain adaptation challenges for image classifiers and will evaluate adaptation to novel viewpoints, backgrounds, modalities and degradation in quality.  ...  The Visual Domain Adaptation (VisDA) 2021 competition tests models' ability to adapt to novel test distributions and handle distributional shift.  ...  AutoML for Lifelong Machine Learning (NeurIPS'18 competition) addressed concept drift in lifelong learning, which is different from unsupervised domain adaptation.  ... 
arXiv:2107.11011v1 fatcat:2vxs4bp27ndmtct3qycshhnkw4

Visual-based Autonomous Driving Deployment from a Stochastic and Uncertainty-aware Perspective [article]

Lei Tai, Peng Yun, Yuying Chen, Congcong Liu, Haoyang Ye, Ming Liu
2019 arXiv   pre-print
End-to-end visual-based imitation learning has been widely applied in autonomous driving.  ...  Experiments in the Carla navigation benchmark show that our strategy outperforms previous methods, especially in dynamic environments.  ...  For the visual domain adaptation methods, five strategies are compared: • Direct: Directly deploy the control policy in the testing environment without any visual domain adap- tations.  ... 
arXiv:1903.00821v2 fatcat:gv3tlohxz5g47a4vofk2rjhu5y

Universal Semi-Supervised Semantic Segmentation [article]

Tarun Kalluri, Girish Varma, Manmohan Chandraker, C V Jawahar
2019 arXiv   pre-print
In contrast to counterpoints such as fine tuning, joint training or unsupervised domain adaptation, universal semi-supervised segmentation ensures that across all domains: (i) a single model is deployed  ...  To address this, we minimize supervised as well as within and cross-domain unsupervised losses, introducing a novel feature alignment objective based on pixel-aware entropy regularization for the latter  ...  For instance, unsupervised domain adaptation usually does not leverage target domain data to improve source performance.  ... 
arXiv:1811.10323v3 fatcat:gzblvf2f4ndndamvpoxda57asu

Domain Adaptation for Outdoor Robot Traversability Estimation from RGB data with Safety-Preserving Loss [article]

Simone Palazzo, Dario C. Guastella, Luciano Cantelli, Paolo Spadaro, Francesco Rundo, Giovanni Muscato, Daniela Giordano, Concetto Spampinato
2020 arXiv   pre-print
We then enhance the model's capabilities by a) addressing domain shifts through gradient-reversal unsupervised adaptation, and b) accounting for the specific safety requirements of a mobile robot, by encouraging  ...  Being able to estimate the traversability of the area surrounding a mobile robot is a fundamental task in the design of a navigation algorithm.  ...  on "on-road" domain, with unsupervised adaptation on "off-road".  ... 
arXiv:2009.07565v1 fatcat:o3k2rvhrije47mlyxoycr334ga

MotionTransformer: Transferring Neural Inertial Tracking between Domains

Changhao Chen, Yishu Miao, Chris Xiaoxuan Lu, Linhai Xie, Phil Blunsom, Andrew Markham, Niki Trigoni
2019 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
To overcome the challenges of domain adaptation on long sensory sequences, we propose MotionTransformer - a novel framework that extracts domain-invariant features of raw sequences from arbitrary domains  ...  Inertial information processing plays a pivotal role in egomotion awareness for mobile agents, as inertial measurements are entirely egocentric and not environment dependent.  ...  Lilian Zhang at National University of Defense Technology, China for their useful assistance and valuable discussion, who are supported by the National Natural Science Foundation of China (Grants Nos.  ... 
doi:10.1609/aaai.v33i01.33018009 fatcat:5u4zvhbcoratljherau3ux2rta

VCIP 2020 Index

2020 2020 IEEE International Conference on Visual Communications and Image Processing (VCIP)  
Special Cane with Visual Odometry for Real-tim Indoor Navigation of Blind People Zhai, Guangtao Wearable Visually Assistive Device for Blind People to Appreciate Real-world Scene and Screen Image  ...  Special Cane with Visual Odometry for Real-tim Indoor Navigation of Blind People Zhang, Jian Automatic Sheep Counting by Multi-object Tracking Zhang, Jian Wearable Visually Assistive Device  ... 
doi:10.1109/vcip49819.2020.9301896 fatcat:bdh7cuvstzgrbaztnahjdp5s5y

Domain Adaptation Through Task Distillation [article]

Brady Zhou, Nimit Kalra, Philipp Krähenbühl
2020 arXiv   pre-print
Furthermore, it shows promising results on standard domain adaptation benchmarks.  ...  Our core observation is that for certain tasks, such as image recognition, datasets are plentiful. They exist in any interesting domain, simulated or real, and are easy to label and extend.  ...  Domain Adaptation Through Task Distillation  ... 
arXiv:2008.11911v1 fatcat:bijayemzr5f65o7q2bkumxt6um

Usingtagflakefor condensing navigable tag hierarchies from tag clouds

Luigi Di Caro, K. Selçuk Candan, Maria Luisa Sapino
2008 Proceeding of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining - KDD 08  
, suitable for navigation, visualization, classification, and tracking.  ...  This provides tagFlake with a mechanism for enabling navigation within the tag space and for classification of the text documents based on the contextual structure captured by the created hierarchy. tagFlake  ...  The two dimensions used for visualizing the tag/document space are adaptively selected by tagFlake based on the current navigation context (the dimensions are chosen for the tag "storm" in this example  ... 
doi:10.1145/1401890.1402021 dblp:conf/kdd/CaroCS08 fatcat:5ozpql7xh5dvlhpkwrzvpfss6a
« Previous Showing results 1 — 15 out of 7,073 results