409 Hits in 4.2 sec

Reasoning about Fine-grained Attribute Phrases using Reference Games [article]

Jong-Chyi Su, Chenyun Wu, Huaizu Jiang, Subhransu Maji
2017 arXiv   pre-print
We also show that embedding an image into the semantic space of attribute phrases derived from listeners offers 20% improvement in accuracy over existing attribute-based representations on the FGVC-aircraft  ...  We then learn to describe and ground these phrases to images in the context of a *reference game* between a speaker and a listener.  ...  We show that DS is better than SS, and reranking improves the quality of top sentences. red and blue pointy nose; on the runway red plane; many windows; facing rightFigure 7.  ... 
arXiv:1708.08874v1 fatcat:dlxudfbeqnadzai3bjtxiz4pnm

BUPT & ORANGELABS(OrangeBJ) at TRECVID 2014: Instance Search

Yanchao Feng, Yuan Dong, Yue Wu, Hongliang Bai, Shusheng Cen, Bo Liu, Kai Wang, Yuxuan Liu
2014 TREC Video Retrieval Evaluation  
There are total eleven submissions namely: OrangeBJ.F_X_NO_OrangeBJ_1_A: Color-SIFT feature, BOW, Lucene-based search, and face verification with SET-A.  ...  Two automatic and one interactive programs have been presented with 2014 TRECVID INS dataset, which has 464 hours of the BBC EastEnders programs, available in MPEG-4 format (Programme material copyrighted  ...  audio recognition, image recognition, face recognition etc.  ... 
dblp:conf/trecvid/FengDWBCLWL14 fatcat:yeihq56nr5fkdl2bz6of4jxwe4

Concept-Based Embeddings for Natural Language Processing [article]

Yukun Ma, Erik Cambria
2018 arXiv   pre-print
In a broad context of opinion understanding system, we investigate the use of the fused embedding for several core NLP tasks: named entity detection and classification, automatic speech recognition reranking  ...  In this work, we focus on effectively leveraging and integrating information from concept-level as well as word-level via projecting concepts and words into a lower dimensional space while retaining most  ...  In addition, we found that introducing concept-level knowledge to RBM-based reranker results in a better recognition of content words.  ... 
arXiv:1807.05519v1 fatcat:t4l263zvl5e6dizmoauad55nxq

Indexing billions of images for sketch-based retrieval

Xinghai Sun, Changhu Wang, Chao Xu, Lei Zhang
2013 Proceedings of the 21st ACM international conference on Multimedia - MM '13  
It is clear that the results are greatly enhanced new descriptor for shape matching and object recognition.  ...  only share similar shapes, image search purely based on label the top 20 returned search results for each query and the sketch often leads to ambiguous results.  ... 
doi:10.1145/2502081.2502281 dblp:conf/mm/SunWXZ13 fatcat:zscz247harcc3kxoodyyfs32eq

Video search reranking via information bottleneck principle

Winston H. Hsu, Lyndon S. Kennedy, Shih-Fu Chang
2006 Proceedings of the 14th annual ACM international conference on Multimedia - MULTIMEDIA '06  
The IB reranking method, based on a rigorous Information Bottleneck (IB) principle, finds the optimal clustering of images that preserves the maximal mutual information between the search relevance and  ...  Evaluating the approach on the TRECVID 2003-2005 data sets shows significant improvement upon the text search baseline, with relative increases in average performance of up to 23%.  ...  On the other hand, if we apply IB reranking after story-based text search, we get an improvement in MAP to 0.285, a big improvement of over 23%.  ... 
doi:10.1145/1180639.1180654 dblp:conf/mm/HsuKC06 fatcat:q3asbun4gfgntgotmxioitf4y4

Large Scale Retrieval and Generation of Image Descriptions

Vicente Ordonez, Xufeng Han, Polina Kuznetsova, Girish Kulkarni, Margaret Mitchell, Kota Yamaguchi, Karl Stratos, Amit Goyal, Jesse Dodge, Alyssa Mensch, Hal Daumé, Alexander C. Berg (+2 others)
2015 International Journal of Computer Vision  
What is the relationship between pictures, language, and information we can extract using state of the art computational recognition systems?  ...  The end result is two simple, but effective, methods for harnessing the power of big data to produce image captions that are altogether more general, relevant, and human-like than previous attempts.  ...  We evaluate two possible methods for reranking: 1) PageRank based reranking using visual and/or text similarity, 2) Phrase-level TFIDF based reranking.  ... 
doi:10.1007/s11263-015-0840-y fatcat:o3vyf24rm5g2fnvrbwdpnhpfsi

A bag-of-objects retrieval model for web image search

Yang Yang, Linjun Yang, Gangshan Wu, Shipeng Li
2012 Proceedings of the 20th ACM international conference on Multimedia - MM '12  
Then, the query and its result images are represented as a language model on the query-relevant object vocabulary, based on which the ranking function can be derived.  ...  Image search reranking has been an active research topic in recent years to boost the performance of the existing web image search engine which is mostly based on textual metadata of images.  ...  The first row is the result by text based search and the second row is the result after reranking based on PRF assumption.  ... 
doi:10.1145/2393347.2393362 dblp:conf/mm/YangYWL12 fatcat:if5pvghz5ffftgne2qvgamw5l4

Effective training of convolutional networks using noisy Web images

Phong D. Vo, Alexandru Ginsca, Herve Le Borgne, Adrian Popescu
2015 2015 13th International Workshop on Content-Based Multimedia Indexing (CBMI)  
Results show that our method outperforms the conventional method in all of the three datasets.  ...  We evaluate the proposed training method versus the conventional supervised training on cross-domain classification tasks.  ...  Based on reranking results, several training pipelines are proposed to combine training methodologies (train from scratch or finetuning) and data (without reranking, unsupervised or weaklysupervised reranking  ... 
doi:10.1109/cbmi.2015.7153607 dblp:conf/cbmi/VoGBP15 fatcat:sejc5m4wabczdeoas7wquivkdu

Multimedia search reranking

Tao Mei, Yong Rui, Shipeng Li, Qi Tian
2014 ACM Computing Surveys  
Multimedia search re-ranking, which reorders visual documents based on multimodal cues to improve initial text-only searches, has received increasing attention in recent years.  ...  Such a problem is challenging because the initial search results often have a great deal of noise.  ...  The system then reranks the previous search results based on the user relevance feedback via active learning.  ... 
doi:10.1145/2536798 fatcat:6kmga3jo4fa2tp4354wukfuzja

Face image retrieval based on shape and texture feature fusion

Zongguang Lu, Jing Yang, Qingshan Liu
2017 Computational Visual Media  
Extensive experiments on the CASIA-WebFace, MSRA-CFW, and LFW datasets illustrate the superiority of our method.  ...  Humongous amounts of data bring various challenges to face image retrieval. This paper proposes an efficient method to solve those problems.  ...  Scalable face image retrieval with identity-based quantization and multireference reranking.  ... 
doi:10.1007/s41095-017-0091-7 fatcat:b2vktp6virfjrfcqjepqlyncuy

Introduction to the special section on real-world face recognition

Gang Hua, Ming-Hsuan Yang, Erik Learned-Miller, Yi Ma, Matthew Turk, David J. Kriegman, Thomas S. Huang
2011 IEEE Transactions on Pattern Analysis and Machine Intelligence  
In "Scalable Face Image Retrieval with Identity-Based Quantization and Multireference Reranking," Wu et al. present a scalable face image retrieval system using identity-based quantization to build a visual  ...  In the consumer digital imaging domain, practical face annotation systems are emerging based on existing face recognition technologies.  ... 
doi:10.1109/tpami.2011.182 pmid:26598769 fatcat:6okfn7ltxfh7tb5vsnbt7okt2i

A Survey of Major Techniques of Clothing in Vision

Dongyan Huang
2015 Journal of Fiber Bioengineering and Informatics  
Firstly, we provide an overview of the techniques for clothing segmentation and clothing recognition. Then, we introduce and analyze the representative works on clothing retrieval.  ...  There are many remarkable advances on the basis of the predecessors' achievements in the three fields.  ...  And a reranking approach was proposed based on the recognition of the common clothing attributes including types and sleeves.  ... 
doi:10.3993/jfbim00175 fatcat:7zuixkdmijbsxishsqbhh7pubu

Orange Labs Beijing (FTRDBJ) at TRECVID 2013 : Instance Search and Video Semantic Indexing Systems

Hongliang Bai, Yuan Dong, Shusheng Cen, Lezi Wang, Lei Liu, Wei Liu, Yunlong Bian, Chong Huang, Nan Zhao, Bo Liu, Yanchao Feng, Peng Li (+2 others)
2013 TREC Video Retrieval Evaluation  
One interactive and three automatic runs have been submitted, namely: F X NO FTRDBJ 1: SIFT feature, Lucene-based search and face recognition F X NO FTRDBJ 2: SIFT feature and Lucene-based search F X NO  ...  FTRDBJ 3: Learned feature by Convolution Neural Networks (CNNs) I X NO FTRDBJ 4: SIFT feature, Lucene-based search and random walk based relative feedback After experiments in 2013 TRECVID INS dataset  ...  , such as Scale Invariant Feature Transformation (SIFT), Shape Context, Gradient Location and Orientation Histogram, Speeded Up Robust Features(SURF), DAISY.  ... 
dblp:conf/trecvid/BaiDCWLLBHZLFLC13 fatcat:t27xtxpccbccxogv43xna2sjsi

Automatic Image Captioning Methods

It uses predefined templates, which will be filled based on the results of the on-scene element detection.  ...  For reranking phrases, either PageRank based reranking using visual and/or text similarity is used or Phrase-level TFIDF based reranking.  ... 
doi:10.35940/ijitee.g1013.0597s20 fatcat:ru7qyejarrgkxayzlmqtxw4sci

Analysis of Different Techniques for Person Re-Identification: An Assessment

Richa Jain, Prof. Seema Shukla
2019 Zenodo  
Also, person determined from utterly completely different camera views has very important variations on poses and viewpoints.  ...  Person re identification is also a tough analysis topic as a result of partial occlusions, low resolution images and massive illumination changes.  ...  Based on the drawbacks introduce a bi-directional reranking method for getting better results.  ... 
doi:10.5281/zenodo.3605938 fatcat:4xadgmsfq5b35lj2mun5dinxqa
« Previous Showing results 1 — 15 out of 409 results