Entity

Time filter

Source Type


Wang Z.,Nanjing Southeast University | Wang Z.,Key Laboratory of Measurement and Control of Complex Systems of Engineering | Huang R.,Nanjing Southeast University | Huang R.,Key Laboratory of Measurement and Control of Complex Systems of Engineering | And 5 more authors.
Chinese Control Conference, CCC | Year: 2014

Feature descriptor based methods (e.g. Local Binary Patterns, Local Ternary Patterns) have gained encouraging results in face recognition. However one needs to manually set the threshold in Local Ternary Patterns (LTP). The threshold in LTP is not data adaptive and not robust to noise. In some cases, we may not give a suitable threshold for LTP. Inspired by Weber's Law, here a data adaptive threshold strategy is prosed for LTP and an enhanced LTP is given for face recognition. We evaluate the enhanced LTP on ORL and FERET face databases and the results demonstrate that the enhanced LTP significantly improves the performances. © 2014 TCCT, CAA. Source


Nguyen H.V.,Nanjing Southeast University | Nguyen H.V.,Nanjing University of Technology | Huang R.,Nanjing Southeast University | Huang R.,Key Laboratory of Measurement and Control of Complex Systems of Engineering | And 5 more authors.
Chinese Control Conference, CCC | Year: 2014

Based on the recent success of Low-Rank matrix Representation (LRR), we propose a novel classification method for robust face recognition, named LRR-based Classification (LRRC). By the ideal that if each data class is linearly spanned by a subspace of unknown dimensions and the data are noiseless, the lowest-rank representations of a set of test vector samples with respect to a set of training vector samples have the nature of being both dense for within-class affinity and almost zero for between-class affinities. Consequently, the LRR exactly reveals the classification of the data. Our experimental results demonstrate that LRRC has competitive with state-of-the-art classification methods. © 2014 TCCT, CAA. Source


Wang L.,Nanjing University of Science and Technology | Wang L.,Carnegie Mellon University | Wang L.,Jiangsu Key Laboratory of Image and Video Understanding for Social Safety | Zhao J.,Carnegie Mellon University | And 5 more authors.
2014 IEEE International Conference on Image Processing, ICIP 2014 | Year: 2014

In this paper, we investigate the problem of weakly supervised object localization in images. For such a problem, the goal is to predict the locations of objects in test images while the labels of the training images are given at image-level. That means a label only indicates whether an image contains objects or not, but does not provide the exact locations of the objects. We propose to address this problem using Maximal Entropy Random Walk (MERW). Specifically, we first train a linear SVM classifier with the weakly labeled data. Based on bag-of-words feature representation, the response of a region to the linear SVM classifier can be formulated as the sum of the feature-weights within the region. For a test image, by properly constructing a graph on the feature-points, the stationary distribution of a MERW can indicate the region with the densest positive feature-weights, and thus provides a probabilistic object localization. Experiments compared with state-of-the-art methods on two datasets validate the performance of our method. © 2014 IEEE. Source

Discover hidden collaborations