原文地址:http://www.shamoxia.com/html/y2011/2856.html
Image Ranking and Retrieval based on Multi-Attribute Queries
IEEE Conference on Computer Vision and Pattern Recognition, (CVPR) 2011,(Oral Presentation)
Behjat Siddiquie, Rogerio S. Feris and Larry S. Davis
[pdf]
作者是马里兰大学的,论文不错,并且还有一篇ICCV2011的文章在申。
“Image Ranking and Retrieval based on Multi-Attribute Queries”这篇文章主要思想是通过一个多属性的查询(例如:戴墨镜擦口红的亚洲长发年轻姑娘),通过把这个查询提交给模型,模型可以综合“墨镜”“口红”“亚洲”“长发”“年轻”“姑娘”这些属性综合进行检索,并且能够学习这些属性之前的关系wp,学习方法用到的是reverse learning的东西,并且和multi-label learning进行集成,优化方法借鉴了structured learning(Thorsten Joachims做了好多structured learning),评测通过ROC面积以及NDCG@100进行的。
效果有改善,感觉不错,不愧是CVPR的Oral,我觉得胜在这个Multi-Attribute Queries点子上了,其他的都是先有的模型和技术。例如求解的时候用的“Bundle Methods for Regularized Risk Minimization” BMRM solver(link),用到的图像特征有:
Color based features include color histograms, color corelograms, color wavelets and color moments. Texture is encoded using wavelet texture and LBP histograms, while shape information is represented using edge histograms, shape moments and SIFT based visual words. To encode spatial information, we extract feature vectors of each feature type from individual grids of five different configurations (Fig. 2) and concatenate them.
BMRW这个程序包也不错,里面有这些解决问题的源码:
- Binary classification
- Hinge [2,7]
- Squared hinge [3]
- Huber-hinge [3]
- Logistic regression [4]
- Exponential [5]
- ROC Score [6]
- Fbeta Score [6]
- Univariate regression
- $\epsilon$-insensitive [8]
- Huber robust [9]
- Least Mean Squares [10]
- Least Absolute Deviation
- Novelty detection (1-class SVM) [11]
- Quantile regression [12]
- Poisson regression [13]
- Ranking
- NDCG (normalized discounted cummulative gain) [14]
- Graph Matching [16]
- Sequence Segmentation and Classification [17]
- Hinge [2,7]
- Squared hinge [3]
- Huber-hinge [3]
- Logistic regression [4]
- Exponential [5]
- ROC Score [6]
- Fbeta Score [6]
- $\epsilon$-insensitive [8]
- Huber robust [9]
- Least Mean Squares [10]
- Least Absolute Deviation
- NDCG (normalized discounted cummulative gain) [14]
===
CVPR 2011关于Image and Video Retrieval的Paper List如下:
805 | 729 | Asymmetric Distances for Binary Embeddings | Albert Gordo (CVC / XRCE); Florent Perronnin; | Poster | Image and Video Retrieval |
1319 | 737 | City-Scale Landmark Identification on Mobile Devices | David Chen (Stanford University); Georges Baatz (Department of Computer Science, ETH Zurich); Kevin Koeser;Sam Tsai;Ramakrishna Vedantham (Nokia Research Center, Palo Alto); Timo Pylvanainen (Nokia Research Center, Tampere); kimmo Roimela (Nokia Research Center); Xin Chen (Navteq); Jeff Bach (Navteq); Marc Pollefeys;Bernd Girod;Radek Grzeszczuk; | Poster | Image and Video Retrieval |
1261 | 745 | Combining attributes and Fisher vectors for efficient image retrieval | Matthijs Douze;Arnau Ramisa (INRIA); Cordelia Schmid; | Poster | Image and Video Retrieval |
859 | 753 | Compact Hashing with Joint Optimization of Search Accuracy and Time | Junfeng He (Columbia University); Regunathan Radhakrishnan;Shih-Fu Chang;Claus Bauer; | Oral | Image and Video Retrieval |
630 | 761 | Edgel Inverted Index for Large-Scale Sketch-based Image Search | Yang Cao (Shanghai Jiao Tong University); Wang Changhu (microsoft); Zhang Liqing (Shanghai Jiao Tong University); Lei Zhang (Microsoft Research Asia); | Poster | Image and Video Retrieval |
670 | 769 | Face Image Retrieval by Shape Manipulation | Brandon Smith (University of Wisconsin-Madiso); Shengqi Zhu (University of Wisconsin-Madison); Li Zhang; | Poster | Image and Video Retrieval |
277 | 777 | Hello neighbor: accurate object retrieval with k-reciprocal nearest neighbors | Qin Danfeng (ETH Zurich); Stephan Gammeter (ETH Zurich); Lukas Bossard (ETH Zurich); Till Quack;Luc VanGool; | Poster | Image and Video Retrieval |
857 | 785 | Hierarchical Semantic Indexing for Large Scale Image Retrieval | Jia Deng (Princeton University); Alexander Berg (Stony Brook University); Li Fei-Fei; | Poster | Image and Video Retrieval |
298 | 793 | Image Annotation via Preferential Random Walk on Bi-relational Graph of Images and Semantic Labels | Hua Wang (Univ. of Texas at Arlington); Heng Huang (University of Texas at Arlington, Department of Computer Science and Engineering); Chris Ding; | Poster | Image and Video Retrieval |
84 | 801 | Image Ranking and Retrieval Based on Multi-Attribute Queries | Behjat Siddiquie (University of Maryland); Rogerio Feris;Larry Davis; | Oral | Image and Video Retrieval |
912 | 809 | Image Retrieval with Geometry-Preserving Visual Phrases | Yimeng Zhang (Cornell University); Zhaoyin Jia (Cornell University); Tsuhan Chen; | Oral | Image and Video Retrieval |
502 | 817 | Iterative Quantization: A Procrustean Approach to Learning Binary Codes | Yunchao Gong (UNC Chapel Hill); Svetlana Lazebnik (University of North Carolina at Chapel Hill); | Oral | Image and Video Retrieval |
1113 | 825 | Learning Image Vicept Description via Mixed-Norm Regularization for Large Scale Semantic Image Search | Liang LI (ICT, CAS); Shuqiang Jiang;Qingming Huang (Graduate Univ of Chinese Academy of Sciences); | Poster | Image and Video Retrieval |
300 | 833 | Learning structured prediction models for interactive image labeling | Thomas Mensink (XRCE); Jakob Verbeek;Gabriela Csurka (XRCE); | Poster | Image and Video Retrieval |
297 | 841 | Locality-Sensitive Support Vector Machine by Exploring Local Feature Correlation for Web Image Annotation | Guo-Jun Qi (ECE, UIUC); Qi Tian (University of Texas at San Antonio); Thomas Huang (UIUC); | Poster | Image and Video Retrieval |
45 | 849 | Noise Resistant Graph Ranking for Improved Web Image Search | Wei Liu (Columbia University); Yu-Gang Jiang (Columbia University); Jiebo Luo;Shih-Fu Chang; | Poster | Image and Video Retrieval |
361 | 857 | Query-Specific Visual Semantic Spaces for Web Image Re-ranking | Ke Liu (CUHK); Xiaogang Wang (The Chinese University of Hong Kong); | Poster | Image and Video Retrieval |
734 | 865 | Random Forest Voting for Fast Action Search | Gang YU (Nanyang technological Univ); Junsong Yuan (Nanyang Technological University); Zicheng Liu; | Poster | Image and Video Retrieval |
1795 | 873 | Random Maximum Margin Hashing | Alexis Joly (INRIA); olivier buisson (ina); | Poster | Image and Video Retrieval |
790 | 881 | Tag Localization with Spatial Correlations and Joint Group Sparsity | Yang Yang (The University of Queensland); Yi Yang (The University of Queensland); Zi Huang (The University of Queensland); Heng Tao Shen (The University of Queensland); Feiping Nie (University of Texas, Arlington); | Poster | Image and Video Retrieval |
1288 | 889 | Total Recall II: Query Expansion Revisited | Ondrej Chum;Andrej Mikulik (CMP, CTU in Prague); Michal Perdoch (CMP, CTU in Prague); Jiri Matas; | Poster | Image and Video Retrieval |
22 | 897 | Towards Cross-Cateogory Knowledge Propagation for Learning Visual Concepts | Guo-Jun Qi (ECE, UIUC); Yong Rui (Microsoft Corporation); Qi Tian (University of Texas at San Antonio); Thomas Huang (UIUC); | Oral | Image and Video Retrieval |
1492 | 905 | Unsupervised Auxiliary Visual Words Discovery for Large-Scale Image Object Retrieval | Yin-Hsi Kuo (National Taiwan University and Academia Sinica); Hsuan-Tien Lin;Wen-Huang Cheng;Yi-Hsuan Yang;Winston Hsu; | Poster | Image and Video Retrieval |