Xiabi Liu

Xin Duan, Xiabi Liu, Xiaopeng Gong et al.

This paper proposes an automatic image co-segmentation algorithm based on deep reinforcement learning (RL). Existing co-segmentation tasks mainly rely on deep learning methods, and the obtained foreground edges are often rough. In order to obtain more precise foreground edges, we use deep RL to solve this problem and achieve the finer segmentation. To our best knowledge, this is the first work to apply RL methods to co-segmentation. We define the problem as a Markov Decision Process (MDP) and optimize it by RL with asynchronous advantage actor-critic (A3C). The RL image co-segmentation network uses the correlation between images to segment common and salient objects from a set of related images. In order to achieve automatic segmentation, our RL-CoSeg method eliminates user's hints. For the image co-segmentation problem, we propose a collaborative RL algorithm based on the A3C model. We propose a Siamese RL co-segmentation network structure to obtain the co-attention of images for co-segmentation. We improve the self-attention for automatic RL algorithm to obtain long-distance dependence and enlarge the receptive field. The image feature information obtained by self-attention can be used to supplement the deleted user's hints and help to obtain more accurate actions. Experimental results have shown that our method can improve the performance effectively on both coarse and fine initial segmentations, and it achieves the state-of-the-art performance on Internet dataset, iCoseg dataset and MLMR-COS dataset.

Xin Duan, Yan Yang, Liyuan Pan et al.

This paper studies co-segmenting the common semantic object in a set of images. Existing works either rely on carefully engineered networks to mine the implicit semantic information in visual features or require extra data (i.e., classification labels) for training. In this paper, we leverage the contrastive language-image pre-training framework (CLIP) for the task. With a backbone segmentation network that independently processes each image from the set, we introduce semantics from CLIP into the backbone features, refining them in a coarse-to-fine manner with three key modules: i) an image set feature correspondence module, encoding global consistent semantic information of the image set; ii) a CLIP interaction module, using CLIP-mined common semantics of the image set to refine the backbone feature; iii) a CLIP regularization module, drawing CLIP towards this co-segmentation task, identifying the best CLIP semantic and using it to regularize the backbone feature. Experiments on four standard co-segmentation benchmark datasets show that the performance of our method outperforms state-of-the-art methods.

Zhoyang Hai, Liyuan Pan, Xiabi Liu et al.

With the concept of teaching being introduced to the machine learning community, a teacher model start using dynamic loss functions to teach the training of a student model. The dynamic intends to set adaptive loss functions to different phases of student model learning. In existing works, the teacher model 1) merely determines the loss function based on the present states of the student model, i.e., disregards the experience of the teacher; 2) only utilizes the states of the student model, e.g., training iteration number and loss/accuracy from training/validation sets, while ignoring the states of the loss function. In this paper, we first formulate the loss adjustment as a temporal task by designing a teacher model with memory units, and, therefore, enables the student learning to be guided by the experience of the teacher model. Then, with a dynamic loss network, we can additionally use the states of the loss to assist the teacher learning in enhancing the interactions between the teacher and the student model. Extensive experiments demonstrate our approach can enhance student learning and improve the performance of various deep models on real-world tasks, including classification, objective detection, and semantic segmentation scenarios.

Weihua Liu, Said Boumaraf, Jianwu Li et al.

Natural gradient descent (NGD) is a powerful optimization technique for machine learning, but the computational complexity of the inverse Fisher information matrix limits its application in training deep neural networks. To overcome this challenge, we propose a novel optimization method for training deep neural networks called structured natural gradient descent (SNGD). Theoretically, we demonstrate that optimizing the original network using NGD is equivalent to using fast gradient descent (GD) to optimize the reconstructed network with a structural transformation of the parameter matrix. Thereby, we decompose the calculation of the global Fisher information matrix into the efficient computation of local Fisher matrices via constructing local Fisher layers in the reconstructed network to speed up the training. Experimental results on various deep networks and datasets demonstrate that SNGD achieves faster convergence speed than NGD while retaining comparable solutions. Furthermore, our method outperforms traditional GDs in terms of efficiency and effectiveness. Thus, our proposed method has the potential to significantly improve the scalability and efficiency of NGD in deep learning applications. Our source code is available at https://github.com/Chaochao-Lin/SNGD.

Huiyu Li, Xiabi Liu, Said Boumaraf et al.

Small object segmentation, like tumor segmentation, is a difficult and critical task in the field of medical image analysis. Although deep learning based methods have achieved promising performance, they are restricted to the use of binary segmentation mask. Inspired by the rigorous mapping between binary segmentation mask and distance map, we adopt distance map as a novel ground truth and employ a network to fulfill the computation of distance map. Specially, we propose a new segmentation framework that incorporates the existing binary segmentation network and a light weight regression network (dubbed as LR-Net). Thus, the LR-Net can convert the distance map computation into a regression task and leverage the rich information of distance maps. Additionally, we derive a shape-aware loss by employing distance maps as penalty map to infer the complete shape of an object. We evaluated our approach on MICCAI 2017 Liver Tumor Segmentation (LiTS) Challenge dataset and a clinical dataset. Experimental results show that our approach outperforms the classification-based methods as well as other existing state-of-the-arts.

Zhaoning Shi, Meng Xiang, Zhaoyang Hai et al.

Genetic algorithms have played an important role in engineering optimization. Traditional GAs treat each gene separately. However, biophysical studies of gene regulatory networks revealed direct associations between different genes. It inspires us to propose an improvement to GA in this paper, Gene Regulatory Genetic Algorithm (GRGA), which, to our best knowledge, is the first time to utilize relationships among genes for improving GA's accuracy and efficiency. We design a directed multipartite graph encapsulating the solution space, called RGGR, where each node corresponds to a gene in the solution and the edge represents the relationship between adjacent nodes. The edge's weight reflects the relationship degree and is updated based on the idea that the edges' weights in a complete chain as candidate solution with acceptable or unacceptable performance should be strengthened or reduced, respectively. The obtained RGGR is then employed to determine appropriate loci of crossover and mutation operators, thereby directing the evolutionary process toward faster and better convergence. We analyze and validate our proposed GRGA approach in a single-objective multimodal optimization problem, and further test it on three types of applications, including feature selection, text summarization, and dimensionality reduction. Results illustrate that our GARA is effective and promising.

Zhaoyang Hai, Xiabi Liu, Yuchen Ren et al.

The field of meta-learning has seen a dramatic rise in interest in recent years. In existing meta-learning approaches, learning tasks for training meta-models are usually collected from public datasets, which brings the difficulty of obtaining a sufficient number of meta-learning tasks with a large amount of training data. In this paper, we propose a meta-learning approach based on randomly generated meta-learning tasks to obtain a parametric loss for classification learning based on big data. The loss is represented by a deep neural network, called meta-loss network (MLN). To train the MLN, we construct a large number of classification learning tasks through randomly generating training data, validation data, and corresponding ground-truth linear classifier. Our approach has two advantages. First, sufficient meta-learning tasks with large number of training data can be obtained easily. Second, the ground-truth classifier is given, so that the difference between the learned classifier and the ground-truth model can be measured to reflect the performance of MLN more precisely than validation accuracy. Based on this difference, we apply the evolutionary strategy algorithm to find out the optimal MLN. The resultant MLN not only leads to satisfactory learning effects on generated linear classifier learning tasks for testing, but also behaves very well on generated nonlinear classifier learning tasks and various public classification tasks. Our MLN stably surpass cross-entropy (CE) and mean square error (MSE) in testing accuracy and generalization ability. These results illustrate the possibility of achieving satisfactory meta-learning effects using generated learning tasks.

Mengqiao Han, Xiabi Liu, Zhaoyang Hai et al.

Knowledge embedded in the weights of the artificial neural network can be used to improve the network structure, such as in network compression. However, the knowledge is set up by hand, which may not be very accurate, and relevant information may be overlooked. Inspired by how learning works in the mammalian brain, we mine the knowledge contained in the weights of the neural network toward automatic architecture learning in this paper. We introduce a switcher neural network (SNN) that uses as inputs the weights of a task-specific neural network (called TNN for short). By mining the knowledge contained in the weights, the SNN outputs scaling factors for turning off and weighting neurons in the TNN. To optimize the structure and the parameters of TNN simultaneously, the SNN and TNN are learned alternately under the same performance evaluation of TNN using stochastic gradient descent. We test our method on widely used datasets and popular networks in classification applications. In terms of accuracy, we outperform baseline networks and other structure learning methods stably and significantly. At the same time, we compress the baseline networks without introducing any sparse induction mechanism, and our method, in particular, leads to a lower compression rate when dealing with simpler baselines or more difficult tasks. These results demonstrate that our method can produce a more reasonable structure.

Weihua Liu, Xiabi Liu

Natural gradient descent (NGD) provided deep insights and powerful tools to deep neural networks. However the computation of Fisher information matrix becomes more and more difficult as the network structure turns large and complex. This paper proposes a new optimization method whose main idea is to accurately replace the natural gradient optimization by reconstructing the network. More specifically, we reconstruct the structure of the deep neural network, and optimize the new network using traditional gradient descent (GD). The reconstructed network achieves the effect of the optimization way with natural gradient descent. Experimental results show that our optimization method can accelerate the convergence of deep network models and achieve better performance than GD while sharing its computational simplicity.

Xi Liu, Xiabi Liu, Huiyu Li et al.

Image co-segmentation has attracted a lot of attentions in computer vision community. In this paper, we propose a new approach to image co-segmentation through introducing the dense connections into the decoder path of Siamese U-net and presenting a new edge enhanced 3D IOU loss measured over distance maps. Considering the rigorous mapping between the signed normalized distance map (SNDM) and the binary segmentation mask, we estimate the SNDMs directly from original images and use them to determine the segmentation results. We apply the Siamese U-net for solving this problem and improve its effectiveness by densely connecting each layer with subsequent layers in the decoder path. Furthermore, a new learning loss is designed to measure the 3D intersection over union (IOU) between the generated SNDMs and the labeled SNDMs. The experimental results on commonly used datasets for image co-segmentation demonstrate the effectiveness of our presented dense structure and edge enhanced 3D IOU loss of SNDM. To our best knowledge, they lead to the state-of-the-art performance on the Internet and iCoseg datasets.

Mengqiao Han, Xiabi Liu, Zhaoyang Hai et al.

Sparse neural networks are important for achieving better generalization and enhancing computation efficiency. This paper proposes a novel learning approach to obtain sparse fully connected layers in neural networks (NNs) automatically. We design a switcher neural network (SNN) to optimize the structure of the task neural network (TNN). The SNN takes the weights of the TNN as the inputs and its outputs are used to switch the connections of TNN. In this way, the knowledge contained in the weights of TNN is explored to determine the importance of each connection and the structure of TNN consequently. The SNN and TNN are learned alternately with stochastic gradient descent (SGD) optimization, targeting at a common objective. After learning, we achieve the optimal structure and the optimal parameters of the TNN simultaneously. In order to evaluate the proposed approach, we conduct image classification experiments on various network structures and datasets. The network structures include LeNet, ResNet18, ResNet34, VggNet16 and MobileNet. The datasets include MNIST, CIFAR10 and CIFAR100. The experimental results show that our approach can stably lead to sparse and well-performing fully connected layers in NNs.

Zhengwen Li, Xiabi Liu

Deep Metric Learning (DML) is helpful in computer vision tasks. In this paper, we firstly introduce DML into image co-segmentation. We propose a novel Triplet loss for Image Segmentation, called IS-Triplet loss for short, and combine it with traditional image segmentation loss. Different from the general DML task which learns the metric between pictures, we treat each pixel as a sample, and use their embedded features in high-dimensional space to form triples, then we tend to force the distance between pixels of different categories greater than of the same category by optimizing IS-Triplet loss so that the pixels from different categories are easier to be distinguished in the high-dimensional feature space. We further present an efficient triple sampling strategy to make a feasible computation of IS-Triplet loss. Finally, the IS-Triplet loss is combined with 3 traditional image segmentation losses to perform image segmentation. We apply the proposed approach to image co-segmentation and test it on the SBCoseg dataset and the Internet dataset. The experimental result shows that our approach can effectively improve the discrimination of pixels' categories in high-dimensional space and thus help traditional loss achieve better performance of image segmentation with fewer training epochs.

Zhaoyang Hai, Xiabi Liu

This paper proposes a meta-learning approach to evolving a parametrized loss function, which is called Meta-Loss Network (MLN), for training the image classification learning on small datasets. In our approach, the MLN is embedded in the framework of classification learning as a differentiable objective function. The MLN is evolved with the Evolutionary Strategy algorithm (ES) to an optimized loss function, such that a classifier, which optimized to minimize this loss, will achieve a good generalization effect. A classifier learns on a small training dataset to minimize MLN with Stochastic Gradient Descent (SGD), and then the MLN is evolved with the precision of the small-dataset-updated classifier on a large validation dataset. In order to evaluate our approach, the MLN is trained with a large number of small sample learning tasks sampled from FashionMNIST and tested on validation tasks sampled from FashionMNIST and CIFAR10. Experiment results demonstrate that the MLN effectively improved generalization compared to classical cross-entropy error and mean squared error.

Songxiao Yang, Xiabi Liu, Zhongshu Zheng et al.

Current Computer-Aided Diagnosis (CAD) methods mainly depend on medical images. The clinical information, which usually needs to be considered in practical clinical diagnosis, has not been fully employed in CAD. In this paper, we propose a novel deep learning-based method for fusing Magnetic Resonance Imaging (MRI)/Computed Tomography (CT) images and clinical information for diagnostic tasks. Two paths of neural layers are performed to extract image features and clinical features, respectively, and at the same time clinical features are employed as the attention to guide the extraction of image features. Finally, these two modalities of features are concatenated to make decisions. We evaluate the proposed method on its applications to Alzheimer's disease diagnosis, mild cognitive impairment converter prediction and hepatic microvascular invasion diagnosis. The encouraging experimental results prove the values of the image feature extraction guided by clinical features and the concatenation of two modalities of features for classification, which improve the performance of diagnosis effectively and stably.

Weihua Liu, Xiabi Liu, Murong Wang et al.

The popular softmax loss and its recent extensions have achieved great success in the deep learning-based image classification. However, the data for training image classifiers usually has different quality. Ignoring such problem, the correct classification of low quality data is hard to be solved. In this paper, we discover the positive correlation between the feature norm of an image and its quality through careful experiments on various applications and various deep neural networks. Based on this finding, we propose a contraction mapping function to compress the range of feature norms of training images according to their quality and embed this contraction mapping function into softmax loss or its extensions to produce novel learning objectives. The experiments on various classification applications, including handwritten digit recognition, lung nodule classification, face verification and face recognition, demonstrate that the proposed approach is promising to effectively deal with the problem of learning on the data with different quality and leads to the significant and stable improvements in the classification accuracy.

Said Boumaraf, Xiabi Liu, Chokri Ferkous et al.

Mammography remains the most prevalent imaging tool for early breast cancer screening. The language used to describe abnormalities in mammographic reports is based on the breast Imaging Reporting and Data System (BI-RADS). Assigning a correct BI-RADS category to each examined mammogram is a strenuous and challenging task for even experts. This paper proposes a new and effective computer-aided diagnosis (CAD) system to classify mammographic masses into four assessment categories in BI-RADS. The mass regions are first enhanced by means of histogram equalization and then semiautomatically segmented based on the region growing technique. A total of 130 handcrafted BI-RADS features are then extrcated from the shape, margin, and density of each mass, together with the mass size and the patient's age, as mentioned in BI-RADS mammography. Then, a modified feature selection method based on the genetic algorithm (GA) is proposed to select the most clinically significant BI-RADS features. Finally, a back-propagation neural network (BPN) is employed for classification, and its accuracy is used as the fitness in GA. A set of 500 mammogram images from the digital database of screening mammography (DDSM) is used for evaluation. Our system achieves classification accuracy, positive predictive value, negative predictive value, and Matthews correlation coefficient of 84.5%, 84.4%, 94.8%, and 79.3%, respectively. To our best knowledge, this is the best current result for BI-RADS classification of breast masses in mammography, which makes the proposed system promising to support radiologists for deciding proper patient management based on the automatically assigned BI-RADS categories.

Xiabi Liu, Xin Duan

Image co-segmentation is important for its advantage of alleviating the ill-pose nature of image segmentation through exploring the correlation between related images. Many automatic image co-segmentation algorithms have been developed in the last decade, which are investigated comprehensively in this paper. We firstly analyze visual/semantic cues for guiding image co-segmentation, including object cues and correlation cues. Then we describe the traditional methods in three categories of object elements based, object regions/contours based, common object model based. In the next part, deep learning based methods are reviewed. Furthermore, widely used test datasets and evaluation criteria are introduced and the reported performances of the surveyed algorithms are compared with each other. Finally, we discuss the current challenges and possible future directions and conclude the paper. Hopefully, this comprehensive investigation will be helpful for the development of image co-segmentation technique.

Huiyu Li, Xiabi Liu, Said Boumaraf et al.

Automatic segmentation of liver tumors in medical images is crucial for the computer-aided diagnosis and therapy. It is a challenging task, since the tumors are notoriously small against the background voxels. This paper proposes a new three-stage curriculum learning approach for training deep networks to tackle this small object segmentation problem. The learning in the first stage is performed on the whole input to obtain an initial deep network for tumor segmenta-tion. Then the second stage of learning focuses the strength-ening of tumor specific features by continuing training the network on the tumor patches. Finally, we retrain the net-work on the whole input in the third stage, in order that the tumor specific features and the global context can be inte-grated ideally under the segmentation objective. Benefitting from the proposed learning approach, we only need to em-ploy one single network to segment the tumors directly. We evaluated our approach on the 2017 MICCAI Liver Tumor Segmentation challenge dataset. In the experiments, our approach exhibits significant improvement compared with the commonly used cascaded counterpart.