Learning from Ambiguously Labeled Face Images

Learning a classifier from ambiguously labeled face images is challenging since training images are not always explicitly-labeled. For instance, face images of two persons in a news photo are not explicitly labeled by their names in the caption. We propose a Matrix Completion for Ambiguity Resolution (MCar) method for predicting the actual labels from ambiguously labeled images. This step is followed by learning a standard supervised classifier from the disambiguated labels to classify new images. To prevent the majority labels from dominating the result of MCar, we generalize MCar to a weighted MCar (WMCar) that handles label imbalance. Since WMCar outputs a soft labeling vector of reduced ambiguity for each instance, we can iteratively refine it by feeding it as the input to WMCar. Nevertheless, such an iterative implementation can be affected by the noisy soft labeling vectors, and thus the performance may degrade. Our proposed Iterative Candidate Elimination (ICE) procedure makes the iterative ambiguity resolution possible by gradually eliminating a portion of least likely candidates in ambiguously labeled faces. We further extend MCar to incorporate the labeling constraints among instances when such prior knowledge is available. Compared to existing methods, our approach demonstrates improvements on several ambiguously labeled datasets.

[1]  Dong Xu,et al.  Learning by Associating Ambiguously Labeled Images , 2013, 2013 IEEE Conference on Computer Vision and Pattern Recognition.

[2]  Cordelia Schmid,et al.  Multiple Instance Metric Learning from Automatically Labeled Bags of Faces , 2010, ECCV.

[3]  Yee Whye Teh,et al.  Names and faces in the news , 2004, CVPR 2004.

[4]  Mahendra Sahare,et al.  A Review of Multi-Class Classification for Imbalanced Data , 2012 .

[5]  Alexander C. Berg,et al.  Who's In the Picture , 2004, NIPS 2004.

[6]  G. Sapiro,et al.  A collaborative framework for 3D alignment and classification of heterogeneous subvolumes in cryo-electron tomography. , 2013, Journal of structural biology.

[7]  Yu-Chiang Frank Wang,et al.  Low-rank matrix recovery with structural incoherence for robust face recognition , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[8]  Xu-Ying Liu,et al.  Towards Class-Imbalance Aware Multi-Label Learning , 2015, IEEE Transactions on Cybernetics.

[9]  Dong Xu,et al.  Automatic Face Naming by Learning Discriminative Affinity Matrices From Weakly Labeled Images , 2015, IEEE Transactions on Neural Networks and Learning Systems.

[10]  Fei Yu,et al.  Disambiguation-Free Partial Label Learning , 2017, IEEE Trans. Knowl. Data Eng..

[11]  Nello Cristianini,et al.  Controlling the Sensitivity of Support Vector Machines , 1999 .

[12]  Yi Ma,et al.  Robust principal component analysis? , 2009, JACM.

[13]  Ben Taskar,et al.  Learning from Partial Labels , 2011, J. Mach. Learn. Res..

[14]  Thomas G. Dietterich,et al.  Learnability of the Superset Label Learning Problem , 2014, ICML.

[15]  Alexandre Bernardino,et al.  Matrix Completion for Weakly-Supervised Multi-Label Image Classification , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[16]  Yi Lin,et al.  Support Vector Machines for Classification in Nonstandard Situations , 2002, Machine Learning.

[17]  Bao-Liang Lu,et al.  A part-versus-part method for massively parallel training of support vector machines , 2004, 2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541).

[18]  B. Taskar,et al.  Learning from ambiguously labeled images , 2009, 2009 IEEE Conference on Computer Vision and Pattern Recognition.

[19]  Eyke Hüllermeier,et al.  Learning from ambiguously labeled examples , 2005, Intell. Data Anal..

[20]  Tong Zhang,et al.  Statistical Analysis of Some Multi-Category Large Margin Classification Methods , 2004, J. Mach. Learn. Res..

[21]  Erik G. Learned-Miller,et al.  Unsupervised Joint Alignment of Complex Images , 2007, 2007 IEEE 11th International Conference on Computer Vision.

[22]  Jesús Cid-Sueiro,et al.  Proper losses for learning from partial labels , 2012, NIPS.

[23]  Yi Ma,et al.  The Augmented Lagrange Multiplier Method for Exact Recovery of Corrupted Low-Rank Matrices , 2010, Journal of structural biology.

[24]  Fernando De la Torre,et al.  Robust Regression , 2016, IEEE Trans. Pattern Anal. Mach. Intell..

[25]  Francisco Charte,et al.  Addressing imbalance in multilabel classification: Measures and random resampling algorithms , 2015, Neurocomputing.

[26]  Chih-Jen Lin,et al.  LIBSVM: A library for support vector machines , 2011, TIST.

[27]  Gérard Govaert,et al.  Learning from an imprecise teacher : probabilistic and evidential approaches , 2015 .

[28]  Thomas M. Cover,et al.  Geometrical and Statistical Properties of Systems of Linear Inequalities with Applications in Pattern Recognition , 1965, IEEE Trans. Electron. Comput..

[29]  Rong Jin,et al.  Learning with Multiple Labels , 2002, NIPS.

[30]  Thomas G. Dietterich,et al.  A Conditional Multinomial Mixture Model for Superset Label Learning , 2012, NIPS.

[31]  Rama Chellappa,et al.  Matrix completion for resolving label ambiguity , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[32]  Yoram Singer,et al.  Efficient projections onto the l1-ball for learning in high dimensions , 2008, ICML '08.

[33]  Rama Chellappa,et al.  Non-linear dictionary learning with partially labeled data , 2015, Pattern Recognit..

[34]  Haibo He,et al.  Learning from Imbalanced Data , 2009, IEEE Transactions on Knowledge and Data Engineering.

[35]  Andrew Zisserman,et al.  Hello! My name is... Buffy'' -- Automatic Naming of Characters in TV Video , 2006, BMVC.

[36]  Ken Chen,et al.  Efficient Classification of Multi-label and Imbalanced Data using Min-Max Modular Classifiers , 2006, The 2006 IEEE International Joint Conference on Neural Network Proceedings.

[37]  Baoyuan Wu,et al.  Constrained Submodular Minimization for Missing Labels and Class Imbalance in Multi-label Learning , 2016, AAAI.

[38]  Fei Yu,et al.  Solving the Partial Label Learning Problem: An Instance-Based Approach , 2015, IJCAI.

[39]  Alexandre Bernardino,et al.  Matrix Completion for Multi-label Image Classification , 2011, NIPS.

[40]  Rama Chellappa,et al.  Ambiguously Labeled Learning Using Dictionaries , 2014, IEEE Transactions on Information Forensics and Security.

[41]  Zhi-Hua Zhou,et al.  Multi-Instance Multi-Label Learning with Application to Scene Classification , 2006, NIPS.

[42]  E. Candès,et al.  Exact low-rank matrix completion via convex optimization , 2008, 2008 46th Annual Allerton Conference on Communication, Control, and Computing.

[43]  Robert D. Nowak,et al.  Transduction with Matrix Completion: Three Birds with One Stone , 2010, NIPS.

[44]  Francesco Orabona,et al.  Learning from Candidate Labeling Sets , 2010, NIPS.

[45]  Emmanuel J. Candès,et al.  A Singular Value Thresholding Algorithm for Matrix Completion , 2008, SIAM J. Optim..

[46]  Rama Chellappa,et al.  Dictionary Learning from Ambiguously Labeled Data , 2013, 2013 IEEE Conference on Computer Vision and Pattern Recognition.