Mingming Gong

Large-Cone Nonnegative Matrix Factorization

Nonnegative matrix factorization (NMF) has been greatly popularized by its parts-based interpretation and the effective multiplicative updating rule for searching local solutions. In this paper, we study the problem of how to obtain an attractive local solution for NMF, which not only fits the given training data well but also generalizes well on the unseen test data. Based on the geometric interpretation of NMF, we introduce two large-cone penalties for NMF and propose large-cone NMF (LCNMF) algorithms. Compared with NMF, LCNMF will obtain bases comprising a larger simplicial cone, and therefore has three advantages. 1) the empirical reconstruction error of LCNMF could mostly be smaller; (2) the generalization ability of the proposed algorithm is much more powerful; and (3) the obtained bases of LCNMF have a low-overlapping property, which enables the bases to be sparse and makes the proposed algorithms very robust. Experiments on synthetic and real-world data sets confirm the efficiency of LCNMF.

Local Metric Learning for Exemplar-Based Object Detection

Object detection has been widely studied in the computer vision community and it has many real applications, despite its variations, such as scale, pose, lighting, and background. Most classical object detection methods heavily rely on category-based training to handle intra-class variations. In contrast to classical methods that use a rigid category-based representation, exemplar-based methods try to model variations among positives by learning from specific positive samples. However, current existing exemplar-based methods either fail to use any training information or suffer from a significant performance drop when few exemplars are available. In this paper, we design a novel local metric learning approach to well handle exemplar-based object detection task. The main works are two-fold: 1) a novel local metric learning algorithm called exemplar metric learning (EML) is designed and 2) an exemplar-based object detection algorithm based on EML is implemented. We evaluate our method on two generic object detection data sets: UIUC-Car and UMass FDDB. Experiments show that compared with other exemplar-based methods, our approach can effectively enhance object detection performance when few exemplars are available.

DERF: Distinctive Efficient Robust Features From the Biological Modeling of the P Ganglion Cells

Studies in neuroscience and biological vision have shown that the human retina has strong computational power, and its information representation supports vision tasks on both ventral and dorsal pathways. In this paper, a new local image descriptor, termed distinctive efficient robust features (DERF), is derived by modeling the response and distribution properties of the parvocellular-projecting ganglion cells in the primate retina. DERF features exponential scale distribution, exponential grid structure, and circularly symmetric function difference of Gaussian (DoG) used as a convolution kernel, all of which are consistent with the characteristics of the ganglion cell array found in neurophysiology, anatomy, and biophysics. In addition, a new explanation for local descriptor design is presented from the perspective of wavelet tight frames. DoG is naturally a wavelet, and the structure of the grid points array in our descriptor is closely related to the spatial sampling of wavelets. The DoG wavelet itself forms a frame, and when we modulate the parameters of our descriptor to make the frame tighter, the performance of the DERF descriptor improves accordingly. This is verified by designing a tight frame DoG, which leads to much better performance. Extensive experiments conducted in the image matching task on the multiview stereo correspondence data set demonstrate that DERF outperforms state of the art methods for both hand-crafted and learned descriptors, while remaining robust and being much faster to compute.

A method using long digital straight segments for fingerprint recognition

In this paper, we proposed a new method using long digital straight segments (LDSSs) for fingerprint recognition based on such a discovery that LDSSs in fingerprints can accurately characterize the global structure of fingerprints. Different from the estimation of orientation using the slope of the straight segments, the length of LDSSs provides a measure for stability of the estimated orientation. In addition, each digital straight segment can be represented by four parameters: x-coordinate, y-coordinate, slope and length. As a result, only about 600 bytes are needed to store all the parameters of LDSSs of a fingerprint, as is much less than the storage orientation field needs. Finally, the LDSSs can well capture the structural information of local regions. Consequently, LDSSs are more feasible to apply to the matching process than orientation fields. The experiments conducted on fingerprint databases FVC2002 DB3a and DB4a show that our method is effective.

MoE-SPNet: A mixture-of-experts scene parsing network

Abstract Scene parsing is an indispensable component in understanding the semantics within a scene. Traditional methods rely on handcrafted local features and probabilistic graphical models to incorporate local and global cues. Recently, methods based on fully convolutional neural networks have achieved new records on scene parsing. An important strategy common to these methods is the aggregation of hierarchical features yielded by a deep convolutional neural network. However, typical algorithms usually aggregate hierarchical convolutional features via concatenation or linear combination, which cannot sufficiently exploit the diversities of contextual information in multi-scale features and the spatial inhomogeneity of a scene. In this paper, we propose a mixture-of-experts scene parsing network (MoE-SPNet) that incorporates a convolutional mixture-of-experts layer to assess the importance of features from different levels and at different spatial locations. In addition, we propose a variant of mixture-of-experts called the adaptive hierarchical feature aggregation (AHFA) mechanism which can be incorporated into existing scene parsing networks that use skip-connections to fuse features layer-wisely. In the proposed networks, different levels of features at each spatial location are adaptively re-weighted according to the local structure and surrounding contextual information before aggregation. We demonstrate the effectiveness of the proposed methods on two scene parsing datasets including PASCAL VOC 2012 and SceneParse150 based on two kinds of baseline models FCN-8s and DeepLab-ASPP.

Efficient Group Learning with Hypergraph Partition in Multi-task Learning

Recently, wide concern has been aroused in multi-task learning (MTL) area, which assumes that affinitive tasks should own similar parameter representation so that joint learning is both appropriate and reciprocal. Researchers also find that imposing similar parameter representation constraint on dissimilar tasks may be harmful to MTL. However, it’s difficult to determine which tasks are similar. Z Kang et al [1] proposed to simultaneously learn the groups and parameters to address this problem. But the method is inefficient and cannot scale to large data. In this paper, using the property of the parameter matrix, we describe the group learning process as permuting the parameter matrix into a block diagonal matrix, which can be modeled as a hypergraph partition problem. The optimization algorithm scales well to large data. Extensive experiments demonstrate that our method is advantageous over existing MTL methods in terms of accuracy and efficiency.

Clustering Based Image Denoising Using SURE-LET

In this paper, we proposed a novel image denoising method based on clustering using SURE-LET. This method divides the images into several clusters and minimize the Steins unbiased risk estimator (SURE) of each cluster independently, which makes different clusters of pixels were denoised by different threshold functions in the image domain. The proposed method included the traditional SURE-LET as its special case, when the clusters reduced to one. Being more flexible, the proposed method results in smaller SURE and MSE. Experimental results show that the proposed method is effective.

Long digital straight segments for fingerprint matching

In this paper, we proposed a new feature extraction method for fingerprint matching. We find that long digital straight segments (LDSSs) in fingerprints can accurately characterize the global structure of fingerprints. Besides, the orientation information of these segments is more stable compared to orientation fields. It is a useful information for representing the fingerprints. In addition, each digital straight segment can be represented by four parameters: a;-coordinate, y-coordinate, slope and length. It only needs about 600 bytes to store all the parameters of LDSSs of a fingerprint, which is much less than the storage orientation field needs. Finally, the LDSSs can automatically capture the maximum structural information of local regions. Consequently, LDSSs are more convenient to combine with the minutiae based methods than orientation fields. The experiments conducted on fingerprint database FVC2002 DB3a and DB4a show that our method is effective