Strong Representation Power Research Articles

Query-Adaptive Image Retrieval by Deep-Weighted Hashing

Hashing methods have attracted much attention for large scale image retrieval. Some deep hashing methods have achieved promising results by taking advantage of the strong representation power of deep networks recently. However, existing deep hashing methods treat all hash bits equally. On one hand, a large number of images share the same distance to a query image due to the discrete Hamming distance, which raises a critical issue of image retrieval where fine-grained rankings are very important. On the other hand, different hash bits actually contribute to the image retrieval differently, and treating them equally greatly affects the retrieval accuracy of image. To address the above two problems, we propose the query-adaptive deep weighted hashing (QaDWH) approach, which can perform fine-grained ranking for different queries by weighted Hamming distance. First, a novel deep hashing network is proposed to learn the hash codes and corresponding class-wise weights jointly, so that the learned weights can reflect the importance of different hash bits for different image classes. Second, a query-adaptive image retrieval method is proposed, which rapidly generates hash bit weights for different query images by fusing its semantic probability and the learned class-wise weights. Fine-grained image retrieval is then performed by the weighted Hamming distance, which can provide more accurate ranking than the traditional Hamming distance. Experiments on four widely used datasets show that the proposed approach outperforms eight state-of-the-art hashing methods.

Fully Deep Blind Image Quality Predictor

In general, owing to the benefits obtained from original information, full-reference image quality assessment (FR-IQA) achieves relatively higher prediction accuracy than no-reference image quality assessment (NR-IQA). By fully utilizing reference images, conventional FR-IQA methods have been investigated to produce objective scores that are close to subjective scores. In contrast, NR-IQA does not consider reference images; thus, its performance is inferior to that of FR-IQA. To alleviate this accuracy discrepancy between FR-IQA and NR-IQA methods, we propose a blind image evaluator based on a convolutional neural network (BIECON). To imitate FR-IQA behavior, we adopt the strong representation power of a deep convolutional neural network to generate a local quality map, similar to FR-IQA. To obtain the best results from the deep neural network, replacing hand-crafted features with automatically learned features is necessary. To apply the deep model to the NR-IQA framework, three critical problems must be resolved: 1) lack of training data; 2) absence of local ground truth targets; and 3) different purposes of feature learning. BIECON follows the FR-IQA behavior using the local quality maps as intermediate targets for conventional neural networks, which leads to NR-IQA prediction accuracy that is comparable with that of state-of-the-art FR-IQA methods.

CenKNN: a scalable and effective text classifier

A big challenge in text classification is to perform classification on a large-scale and high-dimensional text corpus in the presence of imbalanced class distributions and a large number of irrelevant or noisy term features. A number of techniques have been proposed to handle this challenge with varying degrees of success. In this paper, by combining the strengths of two widely used text classification techniques, K-Nearest-Neighbor (KNN) and centroid based (Centroid) classifiers, we propose a scalable and effective flat classifier, called CenKNN, to cope with this challenge. CenKNN projects high-dimensional (often hundreds of thousands) documents into a low-dimensional (normally a few dozen) space spanned by class centroids, and then uses the \(k\)-d tree structure to find \(K\) nearest neighbors efficiently. Due to the strong representation power of class centroids, CenKNN overcomes two issues related to existing KNN text classifiers, i.e., sensitivity to imbalanced class distributions and irrelevant or noisy term features. By working on projected low-dimensional data, CenKNN substantially reduces the expensive computation time in KNN. CenKNN also works better than Centroid since it uses all the class centroids to define similarity and works well on complex data, i.e., non-linearly separable data and data with local patterns within each class. A series of experiments on both English and Chinese, benchmark and synthetic corpora demonstrates that although CenKNN works on a significantly lower-dimensional space, it performs substantially better than KNN and its five variants, and existing scalable classifiers, including Centroid and Rocchio. CenKNN is also empirically preferable to another well-known classifier, support vector machines, on highly imbalanced corpora with a small number of classes.

Discriminative prototype selection methods for graph embedding

Graphs possess a strong representational power for many types of patterns. However, a main limitation in their use for pattern analysis derives from their difficult mathematical treatment. One way of circumventing this problem is that of transforming the graphs into a vector space by means of graph embedding. Such an embedding can be conveniently obtained by using a set of “prototype” graphs and a dissimilarity measure. However, when we apply this approach to a set of class-labelled graphs, it is challenging to select prototypes capturing both the salient structure within each class and inter-class separation. In this paper, we introduce a novel framework for selecting a set of prototypes from a labelled graph set taking their discriminative power into account. Experimental results showed that such a discriminative prototype selection framework can achieve superior results in classification compared to other well-established prototype selection approaches.