54,616 Hits in 7.4 sec

The Importance of the Depth for Text-Image Selection Strategy in Learning-To-Rank [chapter]

David Buffoni, Sabrina Tollari, Patrick Gallinari
2011 Lecture Notes in Computer Science  
Our work: evaluation of the impact of the depth of a pooling methodology on Learning-to-Rank (LTR) algorithms.  ...  Task: produce a sorted list of images given a user query Problem: how to efficiently learn a ranking function ?  ...  • build lists of top-k documents retrieved by models (BM25, TFIDF and 2 language models) • merge the lists to obtain a training set Our work: what is the influence of k on LTR algorithms in text-image  ... 
doi:10.1007/978-3-642-20161-5_84 fatcat:6vdnk7dravdezocvfwuzbgdnpe


Yi Hsuan Yang, Po Tun Wu, Ching Wei Lee, Kuan Hung Lin, Winston H. Hsu, Homer H. Chen
2008 Proceeding of the 16th ACM international conference on Multimedia - MM '08  
Second, to represent the diversity of search result, we propose an efficient algorithm cannoG to select multiple canonical images without clustering.  ...  First, we propose an ordinal reranking algorithm to enhance the semantic coherence of text-based search result by mining contextual patterns in an unsupervised fashion.  ...  This implies our strategy of selecting images that have high coverage of important context cues is reasonable and practical.  ... 
doi:10.1145/1459359.1459387 dblp:conf/mm/YangWLLHC08 fatcat:s7nthtjrxne4ndwzuebmcyu6cm

Object and Text-guided Semantics for CNN-based Activity Recognition [article]

Sungmin Eum, Christopher Reale, Heesung Kwon, Claire Bonial, Clare Voss
2018 arXiv   pre-print
We further improve upon the multitask learning approach by exploiting a text-guided semantic space to select the most relevant objects with respect to the target activities.  ...  of large text corpora to relate the objects and the activities to be transferred into learning a unified deep convolutional neural network.  ...  Based on an empirical analysis, we selected, for our image input dataset (identified as Y in Figure 2c ), the images that have text-labels for 1000 objects (m = 1000) for training the final version of  ... 
arXiv:1805.01818v1 fatcat:br5ida3bybgqxkn7sh7p5desai

Enhancing Personalized Ads Using Interest Category Classification of SNS Users Based on Deep Neural Networks

Taekeun Hong, Jin-A Choi, Kiho Lim, Pankoo Kim
2020 Sensors  
The results of our extensive experiments show that the classification of users' interests performed best when using texts and images together, at 96.55%, versus texts only, 41.38%, or images only, 93.1%  ...  The classification and recommendation system for identifying social networking site (SNS) users' interests plays a critical role in various industries, particularly advertising.  ...  Conflicts of Interest: The authors declare no conflict of interest.  ... 
doi:10.3390/s21010199 pmid:33396796 fatcat:ff45tf4wazba7ohuqoqmpuna4u

Leveraging user comments for aesthetic aware image search reranking

Jose San Pedro, Tom Yeh, Nuria Oliver
2012 Proceedings of the 21st international conference on World Wide Web - WWW '12  
Text-based query search is the most common approach to retrieve images from the Web.  ...  Two strategies for aesthetic rating inference are proposed: one based on visual content, another based on the analysis of user comments to detect opinions about the quality of images.  ...  In addition, we plan to conduct a large scale qualitative study to determine the reasons behind the preference for different ranking strategies depending on the query.  ... 
doi:10.1145/2187836.2187896 dblp:conf/www/PedroYO12 fatcat:yl5s5opjpvdk7bknw46vmzce4u

Online image classifier learning for Google image search improvement

Yuchai Wan, Xiabi Liu, Jie, Yunpeng Chen
2011 2011 IEEE International Conference on Information and Automation  
Thus it is helpful to reduce the user labor of browsing the ranking in depth for finding the desired images.  ...  When the classifier learning is completed, all the images are re-ranked in descending order of their posterior pseudo-probabilities.  ...  The data shows that our method can reduce the user labor of browsing the ranking in depth for finding the desired images.  ... 
doi:10.1109/icinfa.2011.5948971 fatcat:l4sjpmworre73dlbtyzdo35zyu

Deep Learning for Person Re-identification: A Survey and Outlook [article]

Mang Ye, Jianbing Shen, Gaojie Lin, Tao Xiang, Ling Shao, Steven C. H. Hoi
2021 arXiv   pre-print
We first conduct a comprehensive overview with in-depth analysis for closed-world person Re-ID from three different perspectives, including deep feature representation learning, deep metric learning and  ...  This setting is closer to practical applications under specific scenarios. We summarize the open-world Re-ID in terms of five different aspects.  ...  Training strategy The batch sampling strategy plays an important role in discriminative Re-ID model learning, especially for the triplet loss with hard mining.  ... 
arXiv:2001.04193v2 fatcat:4d3thmsr3va2tnu72nawlu2wxy

Multimodal Fusion for Video Search Reranking

Shikui Wei, Yao Zhao, Zhenfeng Zhu, Nan Liu
2010 IEEE Transactions on Knowledge and Data Engineering  
To offer high accuracy on the top-ranked results, CR-Reranking employs a cross-reference (CR) strategy to fuse multimodal cues.  ...  Therefore, it is crucial for search engines to achieve high accuracy on the top-ranked documents.  ...  Based on these observations, we can give a fairly effective solution for selecting some query-relevant shots, which are important for cluster ranking in our approach.  ... 
doi:10.1109/tkde.2009.145 fatcat:shrno5h34be67ejllilrrt5h7m

Extreme video retrieval

Alexander G. Hauptmann, Wei-Hao Lin, Rong Yan, Jun Yang, Ming-Yu Chen
2006 Proceedings of the 14th annual ACM international conference on Multimedia - MULTIMEDIA '06  
We present an efficient system for video search that maximizes the use of human bandwidth, while at the same time exploiting the machine's ability to learn in real-time from user selected relevant video  ...  In either case, as humans search and find relevant results, the system can invisibly re-rank its previous best guesses using a number of knowledge sources, such as image similarity, text similarity, and  ...  In the long run, we intend to incorporate these insights into the system to further improve the active learning by taking the human weaknesses into account as well as the strengths.  ... 
doi:10.1145/1180639.1180721 dblp:conf/mm/HauptmannLYYC06 fatcat:53ae6kpcdbdjxepgu6wms6xwoa

Water level prediction from social media images with a multi-task ranking approach [article]

P. Chaudhary, S. D'Aronco, J.P. Leitao, K. Schindler, J.D. Wegner
2020 arXiv   pre-print
We demonstrate how to effciently learn a predictor from a small set of annotated water levels and a larger set of weaker annotations that only indicate in which of two images the water level is higher,  ...  We introduce a computer vision system that estimates water depth from social media images taken during flooding events, in order to build flood maps in (near) real-time.  ...  Their approach for extracting flood depth information is based only on regular expression patterns in text, while the images are used only to detect the presence of flooding. Starkley et al.  ... 
arXiv:2007.06749v1 fatcat:c7wjsu45dzbutcaueeopl6aofq

Water level prediction from social media images with a multi-task ranking approach

P. Chaudhary, S. D'Aronco, J.P. Leitão, K. Schindler, J.D. Wegner
2020 ISPRS journal of photogrammetry and remote sensing (Print)  
We demonstrate how to efficiently learn a predictor from a small set of annotated water levels and a larger set of weaker annotations that only indicate in which of two images the water level is higher  ...  We introduce a computer vision system that estimates water depth from social media images taken during flooding events, in order to build flood maps in (near) real-time.  ...  The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.  ... 
doi:10.1016/j.isprsjprs.2020.07.003 fatcat:sqypdksyobdxhcajk2szglby4i

Investigating the document structure as a source of evidence for multimedia fragment retrieval

Mouna Torjmen-Khemakhem, Karen Pinel-Sauvagnat, Mohand Boughanem
2013 Information Processing & Management  
Results showed that structural evidences are of high interest to tune the importance of textual context for multimedia retrieval.  ...  Our goal in this paper is to study the impact, in term of effectiveness, of text position relatively to searched objects.  ...  strategy The challenge of this strategy is to select and rank all relevant multimedia fragments, even if the set of retrieved results contain fragments that overlap (Thorough Retrieval) (Westerveld &  ... 
doi:10.1016/j.ipm.2013.06.001 fatcat:ac4vtvzcorbu7mi33kwuv2wkla

Social image retrieval based on topic diversity

Yaxiong Wang, Li Zhu, Xueming Qian
2021 Multimedia tools and applications  
AbstractImage search re-ranking is one of the most important approaches to enhance the text-based image search results.  ...  The clusters are ranked based on the topic distribution vector and the final retrieval image list is obtained by a greedy selection mechanism based on the estimated relevances.  ...  Learning to re-rank is also a popularly used strategy. Ren et al. [32] learn textual and visual mapping matrices respectively.  ... 
doi:10.1007/s11042-020-10221-z fatcat:7xsbwkuslfe7tmxiruyje2be6y

Evaluation of parameters for combining multiple textual sources of evidence for Web image retrieval using genetic programming

Patrícia Correia Saraiva, João M. B. Cavalcanti, Marcos A. Gonçalves, Katia C. Lage dos Santos, Edleno S. de Moura, Ricardo da S. Torres
2012 Journal of the Brazilian Computer Society  
Web image retrieval is a research area that is receiving a lot of attention in the last few years due to the growing availability of images on the Web.  ...  In this paper we present a study about the usage of genetic programming (GP) to address the problem of image retrieval on the World Wide Web by using textual sources of evidence and textual queries.  ...  We adopted TREC-style pooling of the retrieved images with a pool depth of 30. For each query, we ran the baseline methods and assembled the 30 top ranked images retrieved by each ranking strategy.  ... 
doi:10.1007/s13173-012-0087-1 fatcat:topf3oyxhvg47d334knq4k5ub4

Learning from High-Dimensional and Class-Imbalanced Datasets Using Random Forests

Barbara Pes
2021 Information  
This work attempts to give a contribution to this challenging research area by studying the effectiveness of hybrid learning strategies that involve the integration of feature selection techniques, to  ...  Class imbalance and high dimensionality are two major issues in several real-life applications, e.g., in the fields of bioinformatics, text mining and image classification.  ...  The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.  ... 
doi:10.3390/info12080286 fatcat:2avgiaqkajao7lzvlxesyekfkq
« Previous Showing results 1 — 15 out of 54,616 results