Search

Edge federated learning (FL) is an emerging paradigm that trains a global parametric model from distributed datasets based on wireless communications. This paper proposes a unit-modulus over-the-air computation (UMAirComp) framework to facilitate efficient edge federated learning, which simultaneously uploads local model parameters and updates global model parameters via analog beamforming. The proposed framework avoids sophisticated baseband signal processing, leading to low communication delays and implementation costs. Training loss bounds of UMAirComp FL systems are derived and two low-complexity large-scale optimization algorithms, termed penalty alternating minimization (PAM) and accelerated gradient projection (AGP), are proposed to minimize the nonconvex nonsmooth loss bound. Simulation results show that the proposed UMAirComp framework with PAM algorithm achieves a smaller mean square error of model parameters’ estimation, training loss, and test error compared with other benchmark schemes. Moreover, the proposed UMAirComp framework with AGP algorithm achieves satisfactory performance while reduces the computational complexity by orders of magnitude compared with existing optimization algorithms. Finally, we demonstrate the implementation of UMAirComp in a vehicle-to-everything autonomous driving simulation platform. It is found that autonomous driving tasks are more sensitive to model parameter errors than other tasks since the neural networks for autonomous driving contain sparser model parameters. [ABSTRACT FROM AUTHOR]

In an Internet of Vehicle (IoV) system, federated learning (FL) is a new approach to process real-time vehicle data in a distributed way, which can improve the driving experience and service quality. However, due to the high mobility and uncertainty of vehicles, the existing federated learning protocols are difficult to meet the full requirements from an IoV system, such as efficient resource allocation, high precision learning, and fast convergence of learning algorithm. To solve the above problems, in this paper, we propose a semi-synchronous federated learning (Semi-SynFed) protocol, to improve the performance of machine learning at Internet of Vehicles. In Semi-SynFed, we first select appropriate nodes to participate the aggregation by its computing capacity, network capacity and learning value of training samples. Meanwhile, the dynamic waiting time technique is designed to adjust the server waiting time at each round dynamically, which makes federated learning process work much more efficient. Finally, a dynamic aggregation scheme is designed to aggregate model parameters in an asynchronous way. To evaluate the reliability of Semi-SynFed, we establish a simulation of IoVs based on the real world data. In the comparative experiments, the results show that the Semi-SynFed outperforms the existing federated learning protocols in terms of convergence speed and resource consumption. [ABSTRACT FROM AUTHOR]

Temporal and spatial contexts, characterizing target appearance variations and target-background differences, respectively, are crucial for improving the online adaptive ability and instance-level discriminative ability of object tracking. However, most existing trackers focus on either the temporal context or the spatial context during tracking and have not exploited these contexts simultaneously and effectively. In this paper, we propose a Spatial-TEmporal Memory (STEM) network to exploit these contexts jointly for object tracking. Specifically, we develop a key-value structured memory model equipped with a key-value index-based memory reading mechanism to model the spatial and temporal contexts simultaneously. To update the memory with new target states and ensure the diversity of the memory, we introduce a similarity-aware memory update scheme. In addition, we construct an entropy-guided ensemble strategy to fuse the prediction models based on these two contexts, such that these two contexts can be exploited to estimate the target state jointly. Extensive experimental results on eight challenging datasets, including OTB2015, TC128, UAV123, VOT2018, LaSOT, TrackingNet, GOT-10k, and OxUvA, demonstrate that the proposed method performs favorably against state-of-the-art trackers. [ABSTRACT FROM AUTHOR]

Federated Edge Learning (FEL) enables a massive number of edge devices (e.g. smart phones) to train machine learning models collaboratively. Due to the inherent unreliability of participating edge devices and the unpredictable network conditions, the final model accuracy is largely determined by the communication time between edge devices and the parameter server (PS). However, there exists very limited work to quantify the influence of the communication time in FEL and show how to minimize its negative impacts from a theoretic perspective. In this paper, we are among the first to develop a formal model of the communication time in FEL and its influence on the final model accuracy. In our work, the set of edge devices involved in each global iteration is defined as the ECP (Engaged Client Pool). We model the communication time cost as a function with respect to the response time distribution of individual devices and the ECP size. By incorporating communication time cost into the convergence rate analysis, we propose the ECPA and H-ECPA algorithms to automatically adjust the size of the ECP so as to maximize the model accuracy in both homogeneous and heterogeneous networks. We also analyze how the tail shape of response time affects the convergence rate, and prove that the heavy tail shape can significantly lower the model accuracy. Finally, we conduct extensive experiments with real datasets, and the results confirm the correctness of our analysis and demonstrate the superiority of our proposed algorithms. [ABSTRACT FROM AUTHOR]

In semi-supervised medical image segmentation, most previous works draw on the common assumption that higher entropy means higher uncertainty. In this paper, we investigate a novel method of estimating uncertainty. We observe that, when assigned different misclassification costs in a certain degree, if the segmentation result of a pixel becomes inconsistent, this pixel shows a relative uncertainty in its segmentation. Therefore, we present a new semi-supervised segmentation model, namely, conservative-radical network (CoraNet in short) based on our uncertainty estimation and separate self-training strategy. In particular, our CoraNet model consists of three major components: a conservative-radical module (CRM), a certain region segmentation network (C-SN), and an uncertain region segmentation network (UC-SN) that could be alternatively trained in an end-to-end manner. We have extensively evaluated our method on various segmentation tasks with publicly available benchmark datasets, including CT pancreas, MR endocardium, and MR multi-structures segmentation on the ACDC dataset. Compared with the current state of the art, our CoraNet has demonstrated superior performance. In addition, we have also analyzed its connection with and difference from conventional methods of uncertainty estimation in semi-supervised medical image segmentation. [ABSTRACT FROM AUTHOR]

The peak-to-average power ratio (PAPR) reduction and linearization techniques are both effective methods to improve the efficiency of the transmitter in digital video broadcasting (DVB) systems. Traditional methods deploy the PAPR reduction model and the linearization model, respectively, without considering their mutual influence. Therefore, the joint optimizations of PAPR reduction and linearization techniques are proposed. However, these methods train the PAPR reduction model and the linearization model based on the time-division training method. It is difficult to meet the requirements of multiple objectives. To address this issue, this paper proposes a joint PAPR reduction and digital predistortion (DPD) method using the real-valued neural network (RVNN) for Orthogonal Frequency Division Multiplexing (OFDM) systems. The proposed method jointly trains the PAPR reduction function and the DPD function with multi-objective optimization, to achieve PAPR reduction and linearization simultaneously. Especially, this method unifies the PAPR reduction function and the DPD function into one model based on RVNN, and no extra processing is required at the receiver. Compared with the traditional methods, the experimental results show that the proposed method has superior performance in PAPR, adjacent channel power ratio (ACPR) and bit error rate (BER), while having lower computational complexity. [ABSTRACT FROM AUTHOR]

Many real–world domains involve information naturally represented by graphs, where nodes denote basic patterns while edges stand for relationships among them. The graph neural network (GNN) is a machine learning model capable of directly managing graph–structured data. In the original framework, GNNs are inductively trained, adapting their parameters based on a supervised learning environment. However, GNNs can also take advantage of transductive learning, thanks to the natural way they make information flow and spread across the graph, using relationships among patterns. In this paper, we propose a mixed inductive–transductive GNN model, study its properties and introduce an experimental strategy that allows us to understand and distinguish the role of inductive and transductive learning. The preliminary experimental results show interesting properties for the mixed model, highlighting how the peculiarities of the problems and the data can impact on the two learning strategies. [ABSTRACT FROM AUTHOR]

Diversity is one of the most important properties in image captioning, as it reflects various expressions of important concepts presented in an image. However, the most popular metrics cannot well evaluate the diversity of multiple captions. In this paper, we first propose a metric to measure the diversity of a set of captions, which is derived from latent semantic analysis (LSA), and then kernelize LSA using CIDEr (R. Vedantam et al., 2015) similarity. Compared with mBLEU (R. Shetty et al., 2017), our proposed diversity metrics show a relatively strong correlation to human evaluation. We conduct extensive experiments, finding there is a large gap between the performance of the current state-of-the-art models and human annotations considering both diversity and accuracy; the models that aim to generate captions with higher CIDEr scores normally obtain lower diversity scores, which generally learn to describe images using common words. To bridge this “diversity” gap, we consider several methods for training caption models to generate diverse captions. First, we show that balancing the cross-entropy loss and CIDEr reward in reinforcement learning during training can effectively control the tradeoff between diversity and accuracy of the generated captions. Second, we develop approaches that directly optimize our diversity metric and CIDEr score using reinforcement learning. These proposed approaches using reinforcement learning (RL) can be unified into a self-critical (S. J. Rennie et al., 2017) framework with new RL baselines. Third, we combine accuracy and diversity into a single measure using an ensemble matrix, and then maximize the determinant of the ensemble matrix via reinforcement learning to boost diversity and accuracy, which outperforms its counterparts on the oracle test. Finally, inspired by determinantal point processes (DPP), we develop a DPP selection algorithm to select a subset of captions from a large number of candidate captions. The experimental results show that maximizing the determinant of the ensemble matrix outperforms other methods considerably improving diversity and accuracy. [ABSTRACT FROM AUTHOR]

Geographical topic models have been used to mine geo-tagged documents for topical region and geographical topics, and also have applications in recommendations, user mobility modeling, event detection, etc. Existing studies focus on learning effective geographical topic models while ignoring the efficiency issue. However, it is very expensive to train geographical topic models — it may take days to train a geographical topic model of a small scale on a collection of documents with millions of word tokens. In this paper, we propose the first distributed solution, called ${\sf PGeoTopic}$ PGeoTopic , for training geographical topic models. The proposed solution comprises several novel technical components to increase parallelism, reduce memory requirement, and reduce communication cost. Experiments show that our approach for mining geographical topic models is scalable with both model size and data size on distributed systems. [ABSTRACT FROM AUTHOR]

In this paper, we study the fine-grained categorization problem under the few-shot setting, i.e., each fine-grained class only contains a few labeled examples, termed Fine-Grained Few-Shot classification (FGFS). The core predicament in FGFS is the high intra-class variance yet low inter-class fluctuations in the dataset. In traditional fine-grained classification, the high intra-class variance can be somewhat relieved by conducting the supervised training on the abundant labeled samples. However, with few labeled examples, it is hard for the FGFS model to learn a robust class representation with the significantly higher intra-class variance. Moreover, the inter- and intra-class variance are closely related. The significant intra-class variance in FGFS often aggravates the low inter-class variance issue. To address the above challenges, we propose a Target-Oriented Alignment Network (TOAN) to tackle the FGFS problem from both intra- and inter-class perspective. To reduce the intra-class variance, we propose a target-oriented matching mechanism to reformulate the spatial features of each support image to match the query ones in the embedding space. To enhance the inter-class discrimination, we devise discriminative fine-grained features by integrating local compositional concept representations with the global second-order pooling. We conducted extensive experiments on four public datasets for fine-grained categorization, and the results show the proposed TOAN obtains the state-of-the-art. [ABSTRACT FROM AUTHOR]

Decentralized sparsity learning has attracted a significant amount of attention recently due to its rapidly growing applications. To obtain the robust and sparse estimators, a natural idea is to adopt the non-smooth median loss combined with a $\ell _1$ sparsity regularizer. However, most of the existing methods suffer from slow convergence performance caused by the double non-smooth objective. To accelerate the computation, in this paper, we proposed a decentralized surrogate median regression (deSMR) method for efficiently solving the decentralized sparsity learning problem. We show that our proposed algorithm enjoys a linear convergence rate with a simple implementation. We also investigate the statistical guarantee, and it shows that our proposed estimator achieves a near-oracle convergence rate without any restriction on the number of network nodes. Moreover, we establish the theoretical results for sparse support recovery. Thorough numerical experiments and real data study are provided to demonstrate the effectiveness of our method. [ABSTRACT FROM AUTHOR]

Stochastic gradient descent (SGD) has become the method of choice for training highly complex and nonconvex models since it can not only recover good solutions to minimize training errors but also generalize well. Computational and statistical properties are separately studied to understand the behavior of SGD in the literature. However, there is a lacking study to jointly consider the computational and statistical properties in a nonconvex learning setting. In this paper, we develop novel learning rates of SGD for nonconvex learning by presenting high-probability bounds for both computational and statistical errors. We show that the complexity of SGD iterates grows in a controllable manner with respect to the iteration number, which sheds insights on how an implicit regularization can be achieved by tuning the number of passes to balance the computational and statistical errors. As a byproduct, we also slightly refine the existing studies on the uniform convergence of gradients by showing its connection to Rademacher chaos complexities. [ABSTRACT FROM AUTHOR]

This paper analyzes regularization terms proposed recently for improving the adversarial robustness of deep neural networks (DNNs), from a theoretical point of view. Specifically, we study possible connections between several effective methods, including input-gradient regularization, Jacobian regularization, curvature regularization, and a cross-Lipschitz functional. We investigate them on DNNs with general rectified linear activations, which constitute one of the most prevalent families of models for image classification and a host of other machine learning applications. We shed light on essential ingredients of these regularizations and re-interpret their functionality. Through the lens of our study, more principled and efficient regularizations can possibly be invented in the near future. [ABSTRACT FROM AUTHOR]

Predicting the future trajectories of surrounding agents has become an crucial problem to be solved for the safety of autonomous vehicles. Recent studies based on Long Short Term Memory (LSTM) networks have shown powerful abilities to model social interactions. However, many of these approaches focus on spatial interactions of the neighborhood agents but ignore temporal interactions that accompany spatial interactions. In this paper, we propose a Hierarchical Spatio-Temporal Attention architecture (HSTA), which activates the utilization of spatial interactions with different weights, and jointly considers the temporal interactions across time steps of all agents. More specially, the graph attention mechanism (GAT) is presented to capture spatial interactions, the multi-head attention mechanism (MHA) is conducted to encode temporal correlations of interactions and a state gated fusion (SGF) layer is used to integrate spatial and temporal interactions. We evaluate our proposed method against baselines on both pedestrian and vehicle datasets. The results show that our model is effective and achieves state-of-the-art achievements. [ABSTRACT FROM AUTHOR]

Radar radio source (RRS) recognition plays an important role in the fields of military electronic support systems (ESM) and civilian autonomous driving. The rapid development of machine learning technology, especially deep learning, has effectively and efficiently improved RRS intelligent recognition performances when operating in the increasingly complex electromagnetic environment. However, the data sampling limitation and computation cost are still severe challenges in real RRS recognition scenarios. In this paper, we propose a novel network based on meta-transfer learning, called RRSARNet, to achieve effective adaptive RRS recognition in the context of low signal-to-noise ratio (SNR). First, by using the short-time Fourier transform, a six-type small samples RRS simulation dataset with different SNR levels is constructed. Then, a novel RRSARNet, based on metric learning, is proposed, which consists of a four-layer embedding module and a four-layer relational module. Finally, the RRS dataset is divided into training, supporting and testing subsets, which are used to train and test the RRSARNet in a meta-transfer learning method. Experiments on the RRS dataset show that the proposed RRSARNet can achieve an overall accuracy (OA) above 96% and 99% when the SNR is above −15 dB and −10 dB, respectively. Even when the SNR is -30 dB, OA can reach more than 70%. For 5-way 1-shot and 5-way 5-shot experiments, the inference time of an image is about 0.043 and 0.140 milliseconds, respectively. Besides, experiments on the RRS simulation dataset and the two benchmark datasets, the RRSARNet performs better or more competitive than many existing state-of-the-art technologies in terms of recognition accuracy. [ABSTRACT FROM AUTHOR]

Recent workat the U.S. Army CCDC Ground Vehicle Systems Center has shown that machine learning classifiers can quickly construct high-fidelity mobility maps. Training these classifiers, on the other hand, is still a challenge, since each data instance is labeled by performing a computationally intensive, physics-based simulation. In this paper we introduce an active learning framework, based on the query-by-bagging algorithm, that substantially reduces the number of simulations needed to train a classifier. Experimental results suggest that our sampling algorithm can train a neural network, with higher accuracy, using less than half the number of simulations when compared to random sampling. [ABSTRACT FROM AUTHOR]

Automated machine learning (AutoML) has achieved remarkable progress on various tasks, which is attributed to its minimal involvement of manual feature and model designs. However, most of existing AutoML pipelines only touch parts of the full machine learning pipeline, e.g., neural architecture search or optimizer selection. This leaves potentially important components such as data cleaning and model ensemble out of the optimization, and still results in considerable human involvement and suboptimal performance. The main challenges lie in the huge search space assembling all possibilities over all components, as well as the generalization ability over different tasks like image, text, and tabular etc. In this paper, we present a first-of-its-kind fully AutoML pipeline, to comprehensively automate data preprocessing, feature engineering, model generation/selection/training and ensemble for an arbitrary dataset and evaluation metric. Our innovation lies in the comprehensive scope of a learning pipeline, with a novel “life-long” knowledge anchor design to fundamentally accelerate the search over the full search space. Such knowledge anchors record detailed information of pipelines and integrates them with an evolutionary algorithm for joint optimization across components. Experiments demonstrate that the result pipeline achieves state-of-the-art performance on multiple datasets and modalities. Specifically, the proposed framework was extensively evaluated in the NeurIPS 2019 AutoDL challenge, and won the only champion with a significant gap against other approaches, on all the image, video, speech, text and tabular tracks. [ABSTRACT FROM AUTHOR]

Neural architecture search (NAS) has emerged as a promising avenue for automatically designing task-specific neural networks. Existing NAS approaches require one complete search for each deployment specification of hardware or objective. This is a computationally impractical endeavor given the potentially large number of application scenarios. In this paper, we propose Neural Architecture Transfer (NAT) to overcome this limitation. NAT is designed to efficiently generate task-specific custom models that are competitive under multiple conflicting objectives. To realize this goal we learn task-specific supernets from which specialized subnets can be sampled without any additional training. The key to our approach is an integrated online transfer learning and many-objective evolutionary search procedure. A pre-trained supernet is iteratively adapted while simultaneously searching for task-specific subnets. We demonstrate the efficacy of NAT on 11 benchmark image classification tasks ranging from large-scale multi-class to small-scale fine-grained datasets. In all cases, including ImageNet, NATNets improve upon the state-of-the-art under mobile settings (≤ 600M Multiply-Adds). Surprisingly, small-scale fine-grained datasets benefit the most from NAT. At the same time, the architecture search and transfer is orders of magnitude more efficient than existing NAS methods. Overall, experimental evaluation indicates that, across diverse image classification tasks and computational objectives, NAT is an appreciably more effective alternative to conventional transfer learning of fine-tuning weights of an existing network architecture learned on standard datasets. Code is available at https://github.com/human-analysis/neural-architecture-transfer. [ABSTRACT FROM AUTHOR]

When applying machine learning techniques to the Internet of things, aggregating massive amount of data seriously reduce the system efficiency. To tackle this challenge, a distributed learning framework called federated learning has been proposed. Due to the parallel training structure, the performance of federated learning suffers from the straggler effect. In this paper, to mitigate the straggler effect, we propose a novel learning scheme, edge-assisted federated learning (EAFL), which utilizes edge computing to reduce the computational burdens for stragglers in federated learning. It enables stragglers to offload partial computation to the edge server, and leverages the server's idle computing power to assist clients in model training. The offloading data size is optimized to minimize the learning delay of the system. Based on the optimized data size, a threshold-based offloading strategy for EAFL is proposed. Moreover, we extend EAFL to a dynamic scenario where clients may be offline after several update rounds. By grouping clients into different sets, we formulate the new EAFL delay optimization problem and derive the corresponding offloading strategy for the dynamic scenario. Simulation results are presented to show that EAFL has lower system delay than the original federated learning scheme. [ABSTRACT FROM AUTHOR]

Distant supervision is widely used to extract relational facts with automatically labeled datasets to reduce high cost of human annotation. However, current distantly supervised methods suffer from the common problems of word-level and sentence-level noises, which come from a large proportion of irrelevant words in a sentence and inaccurate relation labels for numerous sentences. The problems lead to unacceptable precision in relation extraction and are critical for the success of using distant supervision. In this paper, we propose a novel and robust neural approach to deal with both problems by reducing influences of the multi-granularity noises. Three levels of noises from word, sentence until knowledge type are carefully considered in this work. We first initiate a question-answering based relation extractor (QARE) to remove noisy words in a sentence. Then we use multi-focus multi-instance learning (MMIL) to alleviate the effects of sentence-level noise by utilizing wrongly labeled sentences properly. Finally, to enhance our method against all the noises, we initialize parameters in our method with a priori knowledge learned from the relevant task of entity type classification by transfer learning. Extensive experiments on both existing benchmark and an improved larger dataset demonstrate that our proposed approach remarkably achieves new state-of-the-art performance. [ABSTRACT FROM AUTHOR]

Federated Learning (FL) has recently received a lot of attention for large-scale privacy-preserving machine learning. However, high communication overheads due to frequent gradient transmissions decelerate FL. To mitigate the communication overheads, two main techniques have been studied: (i) local update of weights characterizing the trade-off between communication and computation and (ii) gradient compression characterizing the trade-off between communication and precision. To the best of our knowledge, studying and balancing those two trade-offs jointly and dynamically while considering their impacts on convergence has remained unresolved even though it promises significantly faster FL. In this paper, we first formulate our problem to minimize learning error with respect to two variables: local update coefficients and sparsity budgets of gradient compression who characterize trade-offs between communication and computation/precision, respectively. We then derive an upper bound of the learning error in a given wall-clock time considering the interdependency between the two variables. Based on this theoretical analysis, we propose an enhanced FL scheme, namely Fast FL (FFL), that jointly and dynamically adjusts the two variables to minimize the learning error. We demonstrate that FFL consistently achieves higher accuracies faster than similar schemes existing in the literature. [ABSTRACT FROM AUTHOR]

Performing deep neural network (DNN) inference in real time requires excessive network resources, which poses a big challenge to the resource-limited industrial Internet of things (IIoT) networks. To address the challenge, in this paper, we introduce an end-edge-cloud orchestration architecture, in which the inference task assignment and DNN model placement are flexibly coordinated. Specifically, the DNN models, trained and pre-stored in the cloud, are properly placed at the end and edge to perform DNN inference. To achieve efficient DNN inference, a multi-dimensional resource management problem is formulated to maximize the average inference accuracy while satisfying the strict delay requirements of inference tasks. Due to the mix-integer decision variables, it is difficult to solve the formulated problem directly. Thus, we transform the formulated problem into a Markov decision process which can be solved efficiently. Furthermore, a deep reinforcement learning based resource management scheme is proposed to make real-time optimal resource allocation decisions. Simulation results are provided to demonstrate that the proposed scheme can efficiently allocate the available spectrum, caching, and computing resources, and improve average inference accuracy by 31.4 % compared with the deep deterministic policy gradient benchmark. [ABSTRACT FROM AUTHOR]

In practice, deep neural networks have been found to be vulnerable to various types of noise, such as adversarial examples and corruption. Various adversarial defense methods have accordingly been developed to improve adversarial robustness for deep models. However, simply training on data mixed with adversarial examples, most of these models still fail to defend against the generalized types of noise. Motivated by the fact that hidden layers play a highly important role in maintaining a robust model, this paper proposes a simple yet powerful training algorithm, named Adversarial Noise Propagation (ANP), which injects noise into the hidden layers in a layer-wise manner. ANP can be implemented efficiently by exploiting the nature of the backward-forward training style. Through thorough investigations, we determine that different hidden layers make different contributions to model robustness and clean accuracy, while shallow layers are comparatively more critical than deep layers. Moreover, our framework can be easily combined with other adversarial training methods to further improve model robustness by exploiting the potential of hidden layers. Extensive experiments on MNIST, CIFAR-10, CIFAR-10-C, CIFAR-10-P, and ImageNet demonstrate that ANP enables the strong robustness for deep models against both adversarial and corrupted ones, and also significantly outperforms various adversarial defense methods. [ABSTRACT FROM AUTHOR]

In this paper, we propose a structure-coherent deep feature learning method for face alignment. Unlike most existing face alignment methods which overlook the facial structure cues, we explicitly exploit the relation among facial landmarks to make the detector robust to hard cases such as occlusion and large pose. Specifically, we leverage a landmark-graph relational network to enforce the structural relationships among landmarks. We consider the facial landmarks as structural graph nodes and carefully design the neighborhood to passing features among the most related nodes. Our method dynamically adapts the weights of node neighborhood to eliminate distracted information from noisy nodes, such as occluded landmark point. Moreover, different from most previous works which only tend to penalize the landmarks absolute position during the training, we propose a relative location loss to enhance the information of relative location of landmarks. This relative location supervision further regularizes the facial structure. Our approach considers the interactions among facial landmarks and can be easily implemented on top of any convolutional backbone to boost the performance. Extensive experiments on three popular benchmarks, including WFLW, COFW and 300W, demonstrate the effectiveness of the proposed method. In particular, due to explicit structure modeling, our approach is especially robust to challenging cases resulting in impressive low failure rate on COFW and WFLW datasets. The model and code are publicly available at https://github.com/BeierZhu/Sturcture-Coherency-Face-Alignment [ABSTRACT FROM AUTHOR]

Inspired by the perceived saturation of human visual system, this paper proposes a two-stream hybrid networks to simulate binocular vision for salient object detection (SOD). Each stream in our system consists of unsupervised and supervised methods to form a two-branch module, so as to model the interaction between human intuition and memory. The two-branch module parallel processes visual information with bottom-up and top-down SODs, and output two initial saliency maps. Then a polyharmonic neural network with random-weight (PNNRW) is utilized to fuse two-branch’s perception and refine the salient objects by learning online via multi-source cues. Depend on visual perceptual saturation, we can select optimal parameter of superpixel for unsupervised branch, locate sampling regions for PNNRW, and construct a positive feedback loop to facilitate perception saturated after the perception fusion. By comparing the binary outputs of the two-stream, the pixel annotation of predicted object with high saturation degree could be taken as new training samples. The presented method constitutes a semi-supervised learning framework actually. Supervised branches only need to be pre-trained initial, the system can collect the training samples with high confidence level and then train new models by itself. Extensive experiments show that the new framework can improve performance of the existing SOD methods, that exceeds the state-of-the-art methods in six popular benchmarks. [ABSTRACT FROM AUTHOR]

Human attention is an interactive activity between our visual system and our brain, using both low-level visual stimulus and high-level semantic information. Previous image salient object detection (SOD) studies conduct their saliency predictions via a multitask methodology in which pixelwise saliency regression and segmentation-like saliency refinement are conducted simultaneously. However, this multitask methodology has one critical limitation: the semantic information embedded in feature backbones might be degenerated during the training process. Our visual attention is determined mainly by semantic information, which is evidenced by our tendency to pay more attention to semantically salient regions even if these regions are not the most perceptually salient at first glance. This fact clearly contradicts the widely used multitask methodology mentioned above. To address this issue, this paper divides the SOD problem into two sequential steps. First, we devise a lightweight, weakly supervised deep network to coarsely locate the semantically salient regions. Next, as a postprocessing refinement, we selectively fuse multiple off-the-shelf deep models on the semantically salient regions identified by the previous step to formulate a pixelwise saliency map. Compared with the state-of-the-art (SOTA) models that focus on learning the pixelwise saliency in single images using only perceptual clues, our method aims at investigating the object-level semantic ranks between multiple images, of which the methodology is more consistent with the human attention mechanism. Our method is simple yet effective, and it is the first attempt to consider salient object detection as mainly an object-level semantic reranking problem. [ABSTRACT FROM AUTHOR]

Photorealistic style transfer is a challenging task, which demands the stylized image remains real. Existing methods are still suffering from unrealistic artifacts and heavy computational cost. In this paper, we propose a novel Style-Corpus Constrained Learning (SCCL) scheme to address these issues. The style-corpus with the style-specific and style-agnostic characteristics simultaneously is proposed to constrain the stylized image with the style consistency among different samples, which improves photorealism of stylization output. By using adversarial distillation learning strategy, a simple fast-to-execute network is trained to substitute previous complex feature transforms models, which reduces the computational cost significantly. Experiments demonstrate that our method produces rich-detailed photorealistic images, with 13 ~ 50 times faster than the state-of-the-art method (WCT2). [ABSTRACT FROM AUTHOR]

The existing fusion-based RGB-D salient object detection methods usually adopt the bistream structure to strike a balance in the fusion trade-off between RGB and depth (D). While the D quality usually varies among the scenes, the state-of-the-art bistream approaches are depth-quality-unaware, resulting in substantial difficulties in achieving complementary fusion status between RGB and D and leading to poor fusion results for low-quality D. Thus, this paper attempts to integrate a novel depth-quality-aware subnet into the classic bistream structure in order to assess the depth quality prior to conducting the selective RGB-D fusion. Compared to the SOTA bistream methods, the major advantage of our method is its ability to lessen the importance of the low-quality, no-contribution, or even negative-contribution D regions during RGB-D fusion, achieving a much improved complementary status between RGB and D. Our source code and data are available online at https://github.com/qdu1995/DQSD. [ABSTRACT FROM AUTHOR]

In visual tracking, how to effectively model the target appearance using limited prior information remains an open problem. In this paper, we leverage an ensemble of diverse models to learn manifold representations for robust object tracking. The proposed ensemble framework includes a shared backbone network for efficient feature extraction and multiple head networks for independent predictions. Trained by the shared data within an identical structure, the mutually correlated head models heavily hinder the potential of ensemble learning. To shrink the representational overlaps among multiple models while encouraging the diversity of individual predictions, we propose the model diversity and response diversity regularization terms during training. By fusing these distinctive prediction results via a fusion module, the tracking variance caused by the distractor objects can be largely restrained. Our whole framework is end-to-end trained in a data-driven manner, avoiding the heuristic designs of multiple base models and fusion strategies. The proposed method achieves state-of-the-art results on seven challenging benchmarks while operating in real-time. [ABSTRACT FROM AUTHOR]

While convolutional neural network (CNN) has achieved overwhelming success in various vision tasks, its heavy computational cost and storage overhead limit the practical use on mobile or embedded devices. Recently, compressing CNN models has attracted considerable attention, where pruning CNN filters, also known as the channel pruning, has generated great research popularity due to its high compression rate. In this paper, a new channel pruning framework is proposed, which can significantly reduce the computational complexity while maintaining sufficient model accuracy. Unlike most existing approaches that seek to-be-pruned filters layer by layer, we argue that choosing appropriate layers for pruning is more crucial, which can result in more complexity reduction but less performance drop. To this end, we utilize a long short-term memory (LSTM) to learn the hierarchical characteristics of a network and generate a global network pruning scheme. On top of it, we propose a data-dependent soft pruning method, dubbed Squeeze-Excitation-Pruning (SEP), which does not physically prune any filters but selectively excludes some kernels involved in calculating forward and backward propagations depending on the pruning scheme. Compared with the hard pruning, our soft pruning can better retain the capacity and knowledge of the baseline model. Experimental results demonstrate that our approach still achieves comparable accuracy even when reducing 70.1% Floating-point operation per second (FLOPs) for VGG and 47.5% for Resnet-56. [ABSTRACT FROM AUTHOR]