Fucong Liu’s research while affiliated with China Mobile Group Design Institute Co. Ltd. and other places

Currently, how to deal with the massive garbage produced by various human activities is a hot topic all around the world. In this paper, a preliminary and essential step is to classify the garbage into different categories. However, the mainstream waste classification mode relies heavily on manual work, which consumes a lot of labor and is very inefficient. With the rapid development of deep learning, convolutional neural networks (CNN) have been successfully applied to various application fields. Therefore, some researchers have directly adopted CNNs to classify garbage through their images. However, compared with other images, the garbage images have their own characteristics (such as inter-class similarity, intra-class variance and complex background). Thus, neglecting these characteristics would impair the classification accuracy of CNN. To overcome the limitations of existing garbage image classification methods, a Depth-wise Separable Convolution Attention Module (DSCAM) is proposed in this paper. In DSCAM, the inherent relationships of channels and spatial positions in garbage image features are captured by two attention modules with depth-wise separable convolutions, so that our method could only focus on important information and ignore the interference. Moreover, we also adopt a residual network as the backbone of DSCAM to enhance its discriminative ability. We conduct the experiments on five garbage datasets. The experimental results demonstrate that the proposed method could effectively classify the garbage images and that it outperforms some classical methods.

Artificial intelligence is one of the most popular topics in computer science. Convolutional neural network (CNN), which is an important artificial intelligence deep learning model, has been widely used in many fields. However, training a CNN requires a large amount of labeled data to achieve a good performance but labeling data is a time-consuming and laborious work. Since active learning can effectively reduce the labeling effort, we propose a new intelligent active learning method for deep learning, which is called multi-view active learning based on double-branch network (MALDB). Different from most existing active learning methods, our proposed MALDB first integrates two Bayesian convolutional neural networks (BCNNs) with different structures as two branches of a classifier to learn the effective features for each sample. Then, MALDB performs data analysis on unlabeled dataset and queries the useful unlabeled samples based on different characteristics of two branches to iteratively expand the training dataset and improve the performance of classifier. Finally, MALDB combines multiple level information from multiple hidden layers of BCNNs to further improve the stability of sample selection. The experiments are conducted on five extensively used datasets, Fashion-MNIST, Cifar-10, SVHN, Scene-15 and UIUC-Sports, the experimental results demonstrate the validity of our proposed MALDB.

In recent years, graph-based dimensionality reduction methods became increasingly more significant since they have been successfully applied in various computer vision and machine learning problems. The key point in graph-based dimensionality reduction methods is how to construct an appropriate graph to reflect the underlying distribution of data. However, most existing methods usually consider graph construction and dimensionality reduction as two separate processes. To overcome this limitation, a multiple locality-constrained graph optimization for dimensionality reduction (MLGODR) algorithm is proposed in this paper. The proposed MLGODR possesses two characteristics. First, MLGODR integrates graph optimization and dimensionality reduction into a unified framework. Thus, a graph that characterizes the distribution of input data and a matrix that projects the input data into a lowdimensional subspace can be learned simultaneously. Second, to better exploit the local structure of input data, a locality constraint that adaptively combines multiple distance measurements is introduced into our objective function. Moreover, an effective updating algorithm is also designed to solve the proposed MLGODR. Extensive experiments are performed on four image databases and four UCI datasets. The experimental results demonstrate that our method outperforms the compared approaches in both classification and cluster tasks.

Feature selection is an interesting and challenging task in data analysis process. In this paper, a novel algorithm named Regularized Matrix Factorization Feature Selection (RMFFS) is proposed for unsupervised feature selection. Compared with other matrix factorization based feature selection methods, a main advantage of our algorithm is that it takes the correlation among features into consideration. Through introducing an inner product regularization into our algorithm, the features selected by RMFFS would not only well represent the original high-dimensional data, but also contain low redundancy. Moreover, a simple yet efficient iteratively updating algorithm is also developed to solve the proposed RMFFS. Extensive experimental results on nine real world databases demonstrate that our proposed method can achieve better performance than some state-of-the-art unsupervised feature selection methods.

Scene recognition is a significant and challenging problem in the field of computer vision. One of the principal bottlenecks in applying machine learning techniques to scene recognition tasks is the requirement of a large number of labeled training data. However, labeling massive training data manually (especially labeling images and videos) is very expensive in terms of human time and effort. In this paper, we present a novel multicriteria-based active discriminative dictionary learning (M-ADDL) algorithm to reduce the human annotation effort and create a robust scene recognition model. The M-ADDL algorithm possesses three advantages. First, M-ADDL introduces an active learning strategy into the discriminative dictionary learning model so that the performance of discriminative dictionary learning can be improved when the number of labeled samples is small. Second, different from most existing active learning methods that measure either the informativeness or representativeness of unlabeled samples to select useful samples for expanding the training dataset, M-ADDL employs both informativeness and representativeness to query useful unlabeled samples, and utilizes the manifold-preserving ability of unlabeled samples as an additional sample selection criterion. Finally, a more effective representativeness criterion is presented based on the reconstruction coefficients of the samples. The experimental results of four standard scene recognition databases demonstrate the feasibility and validity of the proposed M-ADDL algorithm. OAPA