The Experts below are selected from a list of 327 Experts worldwide ranked by ideXlab platform
Fernando De La Torre - One of the best experts on this subject based on the ideXlab platform.
-
Semantic Component analysis
International Conference on Computer Vision, 2015Co-Authors: Calvin Murdock, Fernando De La TorreAbstract:Unsupervised and weakly-supervised visual learning in large image collections are critical in order to avoid the time-consuming and error-prone process of manual labeling. Standard approaches rely on methods like multiple-instance learning or graphical models, which can be computationally intensive and sensitive to initialization. On the other hand, simpler Component analysis or clustering methods usually cannot achieve meaningful invariances or Semantic interpretability. To address the issues of previous work, we present a simple but effective method called Semantic Component Analysis (SCA), which provides a decomposition of images into Semantic Components. Unsupervised SCA decomposes additive image representations into spatially-meaningful visual Components that naturally correspond to object categories. Using an overcomplete representation that allows for rich instance-level constraints and spatial priors, SCA gives improved results and more interpretable Components in comparison to traditional matrix factorization techniques. If weakly-supervised information is available in the form of image-level tags, SCA factorizes a set of images into Semantic groups of superpixels. We also provide qualitative connections to traditional methods for Component analysis (e.g. Grassmann averages, PCA, and NMF). The effectiveness of our approach is validated through synthetic data and on the MSRC2 and Sift Flow datasets, demonstrating competitive results in unsupervised and weakly-supervised Semantic segmentation.
-
ICCV - Semantic Component Analysis
2015 IEEE International Conference on Computer Vision (ICCV), 2015Co-Authors: Calvin Murdock, Fernando De La TorreAbstract:Unsupervised and weakly-supervised visual learning in large image collections are critical in order to avoid the time-consuming and error-prone process of manual labeling. Standard approaches rely on methods like multiple-instance learning or graphical models, which can be computationally intensive and sensitive to initialization. On the other hand, simpler Component analysis or clustering methods usually cannot achieve meaningful invariances or Semantic interpretability. To address the issues of previous work, we present a simple but effective method called Semantic Component Analysis (SCA), which provides a decomposition of images into Semantic Components. Unsupervised SCA decomposes additive image representations into spatially-meaningful visual Components that naturally correspond to object categories. Using an overcomplete representation that allows for rich instance-level constraints and spatial priors, SCA gives improved results and more interpretable Components in comparison to traditional matrix factorization techniques. If weakly-supervised information is available in the form of image-level tags, SCA factorizes a set of images into Semantic groups of superpixels. We also provide qualitative connections to traditional methods for Component analysis (e.g. Grassmann averages, PCA, and NMF). The effectiveness of our approach is validated through synthetic data and on the MSRC2 and Sift Flow datasets, demonstrating competitive results in unsupervised and weakly-supervised Semantic segmentation.
Calvin Murdock - One of the best experts on this subject based on the ideXlab platform.
-
Semantic Component analysis
International Conference on Computer Vision, 2015Co-Authors: Calvin Murdock, Fernando De La TorreAbstract:Unsupervised and weakly-supervised visual learning in large image collections are critical in order to avoid the time-consuming and error-prone process of manual labeling. Standard approaches rely on methods like multiple-instance learning or graphical models, which can be computationally intensive and sensitive to initialization. On the other hand, simpler Component analysis or clustering methods usually cannot achieve meaningful invariances or Semantic interpretability. To address the issues of previous work, we present a simple but effective method called Semantic Component Analysis (SCA), which provides a decomposition of images into Semantic Components. Unsupervised SCA decomposes additive image representations into spatially-meaningful visual Components that naturally correspond to object categories. Using an overcomplete representation that allows for rich instance-level constraints and spatial priors, SCA gives improved results and more interpretable Components in comparison to traditional matrix factorization techniques. If weakly-supervised information is available in the form of image-level tags, SCA factorizes a set of images into Semantic groups of superpixels. We also provide qualitative connections to traditional methods for Component analysis (e.g. Grassmann averages, PCA, and NMF). The effectiveness of our approach is validated through synthetic data and on the MSRC2 and Sift Flow datasets, demonstrating competitive results in unsupervised and weakly-supervised Semantic segmentation.
-
ICCV - Semantic Component Analysis
2015 IEEE International Conference on Computer Vision (ICCV), 2015Co-Authors: Calvin Murdock, Fernando De La TorreAbstract:Unsupervised and weakly-supervised visual learning in large image collections are critical in order to avoid the time-consuming and error-prone process of manual labeling. Standard approaches rely on methods like multiple-instance learning or graphical models, which can be computationally intensive and sensitive to initialization. On the other hand, simpler Component analysis or clustering methods usually cannot achieve meaningful invariances or Semantic interpretability. To address the issues of previous work, we present a simple but effective method called Semantic Component Analysis (SCA), which provides a decomposition of images into Semantic Components. Unsupervised SCA decomposes additive image representations into spatially-meaningful visual Components that naturally correspond to object categories. Using an overcomplete representation that allows for rich instance-level constraints and spatial priors, SCA gives improved results and more interpretable Components in comparison to traditional matrix factorization techniques. If weakly-supervised information is available in the form of image-level tags, SCA factorizes a set of images into Semantic groups of superpixels. We also provide qualitative connections to traditional methods for Component analysis (e.g. Grassmann averages, PCA, and NMF). The effectiveness of our approach is validated through synthetic data and on the MSRC2 and Sift Flow datasets, demonstrating competitive results in unsupervised and weakly-supervised Semantic segmentation.
Yuan Yuan - One of the best experts on this subject based on the ideXlab platform.
-
vehicle detection based on Semantic Component analysis
International Conference on Internet Multimedia Computing and Service, 2014Co-Authors: Guosheng Cui, Qi Wang, Yuan YuanAbstract:Vehicle detection is a hot topic in traffic monitoring applications. Though many researchers has done a lot work towards this direction, the detection in occluded conditions is rarely explored and it still remains a challenge. In this work, we focus on the occlusion problem in vehicle detection and propose a novel method based on Semantic Component analysis and scale consideration. Two contributions are claimed in this procedure: 1) Tackling vehicle detection by Semantic Component detection and synthesis. 2) Addressing the scale variation of vehicles by simple yet effective standard Component definition. The experimental results on two typical surveillance videos show that the proposed method can effectively detect the vehicles in the crowded traffic conditions with occlusion.
-
ICIMCS - Vehicle Detection Based on Semantic Component Analysis
Proceedings of International Conference on Internet Multimedia Computing and Service - ICIMCS '14, 2014Co-Authors: Guosheng Cui, Qi Wang, Yuan YuanAbstract:Vehicle detection is a hot topic in traffic monitoring applications. Though many researchers has done a lot work towards this direction, the detection in occluded conditions is rarely explored and it still remains a challenge. In this work, we focus on the occlusion problem in vehicle detection and propose a novel method based on Semantic Component analysis and scale consideration. Two contributions are claimed in this procedure: 1) Tackling vehicle detection by Semantic Component detection and synthesis. 2) Addressing the scale variation of vehicles by simple yet effective standard Component definition. The experimental results on two typical surveillance videos show that the proposed method can effectively detect the vehicles in the crowded traffic conditions with occlusion.
Stanislas Dehaene - One of the best experts on this subject based on the ideXlab platform.
-
acquisition and processing of an artificial mini language combining Semantic and syntactic elements
Cognition, 2019Co-Authors: Fosca Al Roumi, Dror Dotan, Tianming Yang, Liping Wang, Stanislas DehaeneAbstract:Most artificial grammar tasks require the learning of sequences devoid of meaning. Here, we introduce a learning task that allows studying the acquisition and processing of a mini-language of arithmetic with both syntactic and Semantic Components. In this language, symbols have values that predict the probability of being rewarded for a right or left response. Novel to our paradigm is the presence of a syntactic operator which changes the sign of the subsequent value. By continuously tracking finger movement as participants decided whether to press left or right, we revealed the successive cognitive stages associated with the sequential processing of the Semantic and syntactic elements of this mini-language. All participants were able to understand the Semantic Component, but only half of them learned the rule associated with the syntactic operator. Our results provide an encouraging first step in elucidating the way in which humans acquire non-verbal syntactic structures and show how the finger tracking methodology can shed light on real-time artificial language processing.
Maxym Sjachyn - One of the best experts on this subject based on the ideXlab platform.
-
Semantic Component selection
2009Co-Authors: Maxym SjachynAbstract:The means of locating information quickly and efficiently is a growing area of research. However the real challenge is not related to locating bits of information, but finding those that are relevant. Relevant information resides within unstructured ‘natural’ text. However, understanding natural text and judging information relevancy is a challenge. The challenge is partially addressed by use of Semantic models and reasoning approaches that allow categorisation and (within limited fashion) provide understanding of this information. Nevertheless, many such methods are dependent on expert input and, consequently, are expensive to produce and do not scale. Although automated solutions exist, thus far, these have not been able to approach accuracy levels achievable through use of expert input. This thesis presents SemaCS - a novel nondomain specific automated framework of categorising and searching natural text. SemaCS does not rely on expert input; it is based on actual data being searched and statistical Semantic distances between words. These Semantic distances are used to perform basic reasoning and Semantic query interpretation. The approach was tested through a feasibility study and two case studies. Based on reasoning and analyses of data collected through these studies, it can be concluded that SemaCS provides a domain independent approach of Semantic model generation and query interpretation without expert input. Moreover, SemaCS can be further extended to provide a scalable solution applicable to large datasets (i.e. World Wide Web). This thesis contributes to the current body of knowledge by establishing, adapting, and using novel techniques to define a generic selection/categorisation framework. Implementing the framework outlined in the thesis improves an existing algorithm of Semantic distance acquisition. Finally, as a novel approach to the extraction of Semantic information is proposed, there exists a positive impact on Information Retrieval domain and, specifically, on Natural Language Processing, word disambiguation and Web/Intranet search.
-
Semantic Component selection semacs
Fifth International Conference on Commercial-off-the-Shelf (COTS)-Based Software Systems (ICCBSS'05), 2006Co-Authors: Maxym Sjachyn, Ljerka BeusdukicAbstract:In Component based software development, project success or failure largely depends on correct software Component evaluation. All available evaluation methods require time to analyse Components. Due to the black box nature of Components, preliminary judgments are made based on vendor descriptions. As there is no standard way of describing Components, descriptions have to be interpreted using Semantics and domain knowledge. This paper presents a semi-automated generic method for Component identification and classification based on generic domain taxonomy and user generated Semantic input. Every query is Semantically tailored to what is being looked for, arriving at better results then it is currently possible using available automated categorisation systems.
-
ICCBSS - Semantic Component selection - SemaCS
Fifth International Conference on Commercial-off-the-Shelf (COTS)-Based Software Systems (ICCBSS'05), 1Co-Authors: Maxym Sjachyn, Ljerka Beus-dukicAbstract:In Component based software development, project success or failure largely depends on correct software Component evaluation. All available evaluation methods require time to analyse Components. Due to the black box nature of Components, preliminary judgments are made based on vendor descriptions. As there is no standard way of describing Components, descriptions have to be interpreted using Semantics and domain knowledge. This paper presents a semi-automated generic method for Component identification and classification based on generic domain taxonomy and user generated Semantic input. Every query is Semantically tailored to what is being looked for, arriving at better results then it is currently possible using available automated categorisation systems.