Filters








7,669 Hits in 4.9 sec

Domain and View-point Agnostic Hand Action Recognition [article]

Alberto Sabater, Iñigo Alonso, Luis Montesano, Ana C. Murillo
2021 arXiv   pre-print
This work introduces a novel skeleton-based hand motion representation model that tackles this problem. The framework we propose is agnostic to the application domain or camera recording view-point.  ...  And, more importantly, when performing hand action recognition for action domains and camera perspectives which our approach has not been trained for (cross-domain action classification), our proposed  ...  Up to our knowledge, generalization to unseen hand view-points and domains is still to be studied.  ... 
arXiv:2103.02303v3 fatcat:2gjitbedjfaurbotqjqb2xmhom

AdapNet: Adaptability Decomposing Encoder-Decoder Network for Weakly Supervised Action Recognition and Localization [article]

Xiao-Yu Zhang, Changsheng Li, Haichao Shi, Xiaobin Zhu, Peng Li, Jing Dong
2019 arXiv   pre-print
This paper proposes a novel adaptability decomposing encoder-decoder network to transfer reliable knowledge between trimmed and untrimmed videos for action recognition and localization via bidirectional  ...  As a challenging problem for high-level video understanding, weakly supervised action recognition and localization in untrimmed videos has attracted intensive research attention.  ...  As we know, the domain-adaptable and domain-specific representations depict the video from different point of view.  ... 
arXiv:1911.11961v1 fatcat:qxptpsr5djcd5nnpagxnqvnrpu

DeepGRU: Deep Gesture Recognition Utility [article]

Mehran Maghoumi, Joseph J. LaViola Jr
2019 arXiv   pre-print
We propose DeepGRU, a novel end-to-end deep network model informed by recent developments in deep learning for gesture and action recognition, that is streamlined and device-agnostic.  ...  For instance, we achieve a recognition accuracy of 84.9% and 92.3% on cross-subject and cross-view tests of the NTU RGB+D dataset respectively, and also 100% recognition accuracy on the UT-Kinect dataset  ...  Portions of this research used the NTU RGB+D Action Recognition Dataset [46] made available by the ROSE Lab at the Nanyang Technological University, Singapore.  ... 
arXiv:1810.12514v4 fatcat:wrkdmeczvbfufmvzhk4ty7vscq

Activity, Plan, and Goal Recognition: A Review

Franz A. Van-Horenbeke, Angelika Peer
2021 Frontiers in Robotics and AI  
While action and plan recognition are tasks that humans perform naturally and with little effort, they are still an unresolved problem from the point of view of artificial intelligence.  ...  This review is meant to provide a general view of the problem of activity, plan, and goal recognition as a whole.  ...  the point of view of a higher layer that takes its outputs as elementary actions.  ... 
doi:10.3389/frobt.2021.643010 pmid:34041274 pmcid:PMC8141730 fatcat:hgoy6wjz7rgsxeta5olmpumlj4

Episodic Training for Domain Generalization

Da Li, Jianshu Zhang, Yongxin Yang, Cong Liu, Yi-Zhe Song, Timothy Hospedales
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
Domain generalization (DG) is the challenging and topical problem of learning models that generalize to novel testing domains with different statistics than a set of known training domains.  ...  Furthermore, we consider the pervasive workflow of using an ImageNet trained CNN as a fixed feature extractor for downstream recognition tasks.  ...  The goal is to train an action recognition model on a set of source views (domains), and recognise the action from a novel target view (domain).  ... 
doi:10.1109/iccv.2019.00153 dblp:conf/iccv/LiZYLSH19 fatcat:a4fhygn7cjagrfybv4pboqguwy

Recent Developments in Boolean Matrix Factorization

Pauli Miettinen, Stefan Neumann
2020 Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence  
In the last decade, BMF has received a considerable amount of attention in the data mining and formal concept analysis communities and, more recently, the machine learning and the theory communities also  ...  In this survey, we give a concise summary of the efforts of all of these communities and raise some open questions which in our opinion require further investigation.  ...  Acknowledgements The research leading to these results is partially supported by Accenture LTD and by the Center for Research on Computation and Society at Harvard University.  ... 
doi:10.24963/ijcai.2020/675 dblp:conf/ijcai/KerenGK20 fatcat:v2av6qtykrbn3oai3fyqi7aide

Robotic Pick-and-Place of Novel Objects in Clutter with Multi-Affordance Grasping and Cross-Domain Image Matching [article]

Andy Zeng, Shuran Song, Kuan-Ting Yu, Elliott Donlon, Francois R. Hogan, Maria Bauza, Daolin Ma, Orion Taylor, Melody Liu, Eudald Romo, Nima Fazeli, Ferran Alet (+9 others)
2020 arXiv   pre-print
To achieve this, it first uses a category-agnostic affordance prediction algorithm to select and execute among four different grasping primitive behaviors.  ...  Exhaustive experimental results demonstrate that our multi-affordance grasping achieves high success rates for a wide variety of objects in clutter, and our recognition algorithm achieves high accuracy  ...  /1539099), NVIDIA, and Facebook for hardware, technical, and financial support.  ... 
arXiv:1710.01330v5 fatcat:yyytldcvnfbvnailib366awsg4

Episodic Training for Domain Generalization [article]

Da Li, Jianshu Zhang, Yongxin Yang, Cong Liu, Yi-Zhe Song, Timothy M. Hospedales
2019 arXiv   pre-print
Domain generalization (DG) is the challenging and topical problem of learning models that generalize to novel testing domains with different statistics than a set of known training domains.  ...  Furthermore, we consider the pervasive workflow of using an ImageNet trained CNN as a fixed feature extractor for downstream recognition tasks.  ...  The goal is to train an action recognition model on a set of source views (domains), and recognise the action from a novel target view (domain).  ... 
arXiv:1902.00113v3 fatcat:o5t2vmlmsrfh5mahcwvxncstpy

Robotic pick-and-place of novel objects in clutter with multi-affordance grasping and cross-domain image matching

Andy Zeng, Shuran Song, Kuan-Ting Yu, Elliott Donlon, Francois R. Hogan, Maria Bauza, Daolin Ma, Orion Taylor, Melody Liu, Eudald Romo, Nima Fazeli, Ferran Alet (+9 others)
2019 The international journal of robotics research  
It then executes the action with the highest affordance and recognizes picked objects with a cross-domain image classification framework that matches observed images to product images.  ...  To achieve this, it first uses an object-agnostic grasping framework to map from visual observations to actions: inferring dense pixel-wise probability maps of the affordances for four different grasping  ...  and technical support.  ... 
doi:10.1177/0278364919868017 fatcat:tjzvct4y7rfnfiqk5bhlegpiwi

On the Ethics of Building AI in a Responsible Manner [article]

Shai Shalev-Shwartz, Shaked Shammah, Amnon Shashua
2020 arXiv   pre-print
We argue that a formalism of AI alignment that does not distinguish between strategic and agnostic misalignments is not useful, as it deems all technology as un-safe.  ...  The AI-alignment problem arises when there is a discrepancy between the goals that a human designer specifies to an AI learner and a potential catastrophic outcome that does not reflect what the human  ...  As much as those issues require immediate and focused attention, there is a bigger potential danger at hand of a technology whose ultimate evolutionary end-point could get out of hand and cause havoc on  ... 
arXiv:2004.04644v1 fatcat:cxlid5sj2jdwnkopwhjfetymei

Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity Recognition

Evaggelos Spyrou, Eirini Mathe, Georgios Pikramenos, Konstantinos Kechagias, Phivos Mylonas
2020 Technologies  
In this work, we fill this gap by providing ample experimental results comparing data augmentation and domain adaptation techniques on a cross-viewpoint, human activity recognition task from pose information  ...  This is especially true for video data, and in particular for human activity recognition (HAR) tasks.  ...  under the "Action for the Strategic Development on the Research and Technological Sector", funded by the Operational Programme "Competitiveness, Entrepreneurship and Innovation" (NSRF 2014-2020) and co-financed  ... 
doi:10.3390/technologies8040055 fatcat:k3ooqxpbrjfzvhwpe535wx5ao4

Learning task-agnostic representation via toddler-inspired learning [article]

Kwanyoung Park, Junseok Park, Hyunseok Oh, Byoung-Tak Zhang, Youngki Lee
2021 arXiv   pre-print
Inspired by the toddler's learning procedure, we design an interactive agent that can learn and store task-agnostic visual representation while exploring and interacting with objects in the virtual environment  ...  To tackle this problem, we derive inspiration from a highly intentional learning system via action: the toddler.  ...  For the classification and recognition, we suppose that it is because the agent must recognize and classify the objects to achieve maximal reward, while action of the agent is dependent on the transferred  ... 
arXiv:2101.11221v1 fatcat:d6fhxvc22fclvcmvkndpyqmgoq

Learning Chebyshev Basis in Graph Convolutional Networks for Skeleton-based Action Recognition [article]

Hichem Sahbi
2021 arXiv   pre-print
Extensive experiments, conducted on the challenging task of skeleton-based action recognition, show the generalization ability and the outperformance of our proposed Laplacian design w.r.t. different baselines  ...  filtered signals onto the input graph domain.  ...  RELATED WORK In this section, we discuss the related work both from the methodological and the application point-of-view.  ... 
arXiv:2104.05482v2 fatcat:lyn6gjm36bbwnmr7w4s5x6zfay

Action Recognition with Kernel-based Graph Convolutional Networks [article]

Hichem Sahbi
2020 arXiv   pre-print
Experiments conducted on the challenging task of skeleton-based action recognition show the superiority of the proposed method against different baselines as well as the related work.  ...  and well defined.  ...  Table 2 shows a comparison of action recognition performances (and also runtime per epoch during training), using our KGCN (with different kernels) against standard GCN (referred to as SGCN), shown in  ... 
arXiv:2012.14186v1 fatcat:7jza4ltbczhffk46wajapjbope

Domain-Specific Priors and Meta Learning for Few-Shot First-Person Action Recognition [article]

Huseyin Coskun, Zeeshan Zia, Bugra Tekin, Federica Bogo, Nassir Navab, Federico Tombari, Harpreet Sawhney
2021 arXiv   pre-print
Visual cues we employ include object-object interactions, hand grasps and motion within regions that are a function of hand locations.  ...  This enables transfer of action classification models across public datasets captured with diverse scene and action configurations.  ...  ACKNOWLEDGMENTS The authors would like to thank David Joseph Tan for the valuable discussions and constructive feedback. This work was supported by Microsoft.  ... 
arXiv:1907.09382v2 fatcat:aj7rdwx5ongd7dd2tsk6ybyeuu
« Previous Showing results 1 — 15 out of 7,669 results