Deep learning methods have achieved excellent performance in pose estimation, but the lack of robustness causes the keypoints to change drastically between similar images. In view of this problem, a stable heatmap regression method is proposed to alleviate network vulnerability to small perturbations. We utilize the correlation between different rows and columns in a heatmap to alleviate the multi-peaks problem, and design a highly differentiated heatmap regression to make a keypoint discriminative from surrounding points. A maximum stability training loss is used to simplify the optimization difficulty when minimizing the prediction gap of two similar images. The proposed method achieves a significant advance in robustness over state-of-the-art approaches on two benchmark datasets and maintains high performance.
Image matting refers to the estimation of the opacity of foreground objects. It requires correct contours and fine details of foreground objects for the matting results. To better accomplish human image matting tasks, we propose the Cascade Image Matting Network with Deformable Graph Refinement, which can automatically predict precise alpha mattes from single human images without any additional inputs. We adopt a network cascade architecture to perform matting from low-to-high resolution, which corresponds to coarse-to-fine optimization. We also introduce the Deformable Graph Refinement (DGR) module based on graph neural networks (GNNs) to overcome the limitations of convolutional neural networks (CNNs). The DGR module can effectively capture long-range relations and obtain more global and local information to help produce finer alpha mattes. We also reduce the computation complexity of the DGR module by dynamically predicting the neighbors and apply DGR module to higher--resolution features. Experimental results demonstrate the ability of our CasDGR to achieve state-of-the-art performance on synthetic datasets and produce good results on real human images.
The memorization effect of deep learning hinders its performance to effectively generalize on test set when learning with noisy labels. Prior study has discovered that epistemic uncertainty techniques are robust when trained with noisy labels compared with neural networks without uncertainty estimation. They obtain prolonged memorization effect and better generalization performance under the adversarial setting of noisy labels. Due to its superior performance amongst other selected epistemic uncertainty methods under noisy labels, we focus on Monte Carlo Dropout (MCDropout) and investigate why it is robust when trained with noisy labels. Through empirical studies on datasets MNIST, CIFAR-10, Animal-10n, we deep dive into three aspects of MCDropout under noisy label setting: 1. efficacy: understanding the learning behavior and test accuracy of MCDropout when training set contains artificially generated or naturally embedded label noise; 2. representation volatility: studying the responsiveness of neurons by examining the mean and standard deviation on each neuron's activation; 3. network sparsity: investigating the network support of MCDropout in comparison with deterministic neural networks. Our findings suggest that MCDropout further sparsifies and regularizes the deterministic neural networks and thus provides higher robustness against noisy labels.
Recently, research on explainable recommender systems (RS) has drawn much attention from both academia and industry, resulting in a variety of explainable models. As a consequence, their evaluation approaches vary from model to model, which makes it quite difficult to compare the explainability of different models. To achieve a standard way of evaluating recommendation explanations, we provide three benchmark datasets for EXplanaTion RAnking (denoted as EXTRA), on which explainability can be measured by ranking-oriented metrics. Constructing such datasets, however, presents great challenges. First, user-item-explanation interactions are rare in existing RS, so how to find alternatives becomes a challenge. Our solution is to identify nearly duplicate or even identical sentences from user reviews. This idea then leads to the second challenge, i.e., how to efficiently categorize the sentences in a dataset into different groups, since it has quadratic runtime complexity to estimate the similarity between any two sentences. To mitigate this issue, we provide a more efficient method based on Locality Sensitive Hashing (LSH) that can detect near-duplicates in sub-linear time for a given query. Moreover, we plan to make our code publicly available, to allow other researchers create their own datasets.
Speech enhancement is a task to improve the intelligibility and perceptual quality of degraded speech signal. Recently, neural networks based methods have been applied to speech enhancement. However, many neural network based methods require noisy and clean speech pairs for training. We propose a speech enhancement framework that can be trained with large-scale weakly labelled AudioSet dataset. Weakly labelled data only contain audio tags of audio clips, but not the onset or offset times of speech. We first apply pretrained audio neural networks (PANNs) to detect anchor segments that contain speech or sound events in audio clips. Then, we randomly mix two detected anchor segments containing speech and sound events as a mixture, and build a conditional source separation network using PANNs predictions as soft conditions for speech enhancement. In inference, we input a noisy speech signal with the one-hot encoding of "Speech" as a condition to the trained system to predict enhanced speech. Our system achieves a PESQ of 2.28 and an SSNR of 8.75 dB on the VoiceBank-DEMAND dataset, outperforming the previous SEGAN system of 2.16 and 7.73 dB respectively.
Explaining to users why some items are recommended is critical, as it helps users to make better decisions, increase their satisfaction, and gain their trust in recommender systems (RS). However, existing explainable RS usually consider explanations as side outputs of the recommendation model, which has two problems: (1) it is difficult to evaluate the produced explanations because they are usually model-dependent, and (2) as a result, the possible impacts of those explanations are less investigated. To address the evaluation problem, we propose learning to explain for explainable recommendation. The basic idea is to train a model that selects explanations from a collection as a ranking-oriented task. A great challenge, however, is that the sparsity issue in the user-item-explanation data would be severer than that in traditional user-item relation data, since not every user-item pair can associate with multiple explanations. To mitigate this issue, we propose to perform two sets of matrix factorization by considering the ternary relationship as two groups of binary relationships. To further investigate the impacts of explanations, we extend the traditional item ranking of recommendation to an item-explanation joint-ranking formalization. We study if purposely selecting explanations could achieve certain learning goals, e.g., in this paper, improving the recommendation performance. Experiments on three large datasets verify our solution's effectiveness on both item recommendation and explanation ranking. In addition, our user-item-explanation datasets open up new ways of modeling and evaluating recommendation explanations. To facilitate the development of explainable RS, we will make our datasets and code publicly available.
Millimeter-wave (mmWave) massive multiple-input multiple-output (MIMO) systems rely on large-scale antenna arrays to combat large path-loss at mmWave band. Due to hardware characteristics and deploying environments, mmWave massive MIMO systems are vulnerable to antenna element blockages and failures, which necessitate diagnostic techniques to locate faulty antenna elements for calibration purposes. Current diagnostic techniques require full or partial knowledge of channel state information (CSI), which can be challenging to acquire in the presence of antenna failures. In this letter, we propose a blind diagnostic technique to identify faulty antenna elements in mmWave massive MIMO systems, which does not require any CSI knowledge. By jointly exploiting the sparsity of mmWave channel and failure, we first formulate the diagnosis problem as a joint sparse recovery problem. Then, the atomic norm is introduced to induce the sparsity of mmWave channel over continuous Fourier dictionary. An efficient algorithm based on alternating direction method of multipliers (ADMM) is proposed to solve the proposed problem. Finally, the performance of proposed technique is evaluated through numerical simulations.
Intelligent reflecting surface (IRS) is a promising technology for enhancing wireless communication systems, which adaptively configures massive passive reflecting elements to control wireless channel in a desirable way. Due to hardware characteristics and deploying environments, the IRS may be subject to reflecting element blockages and failures, and hence developing diagnostic techniques is of great significance to system monitoring and maintenance. In this paper, we develop diagnostic techniques for IRS systems to locate faulty reflecting elements and retrieve failure parameters. Three cases of the channel state information (CSI) availability are considered. In the first case where full CSI is available, a compressed sensing based diagnostic technique is proposed, which significantly reduces the required number of measurements. In the second case where only partial CSI is available, we jointly exploit the sparsity of the millimeter-wave channel and the failure, and adopt compressed sparse and low-rank matrix recovery algorithm to decouple channel and failure. In the third case where no CSI is available, a novel atomic norm is introduced as the sparsity-inducing norm of the cascaded channel, and the diagnosis problem is formulated as a joint sparse recovery problem. Finally, proposed diagnostic techniques are validated through numerical simulations.
Pathological images may have large variabilities in color intensities due to inconsistencies in staining process, operator ability, and scanner specifications. These variations hamper the performance of computer-aided diagnosis (CAD) systems. Stain normalization has been used to reduce the color variability and increase the prediction accuracy. However, the conventional methods estimate stain parameters from one single reference image, and the current deep learning based methods have a low computational efficiency and risk to introduce artifacts. In this paper, a fast and robust stain normalization network with only 1.28K parameters named StainNet is proposed. StainNet can learn the color mapping relationship from the whole dataset and adjust the color value in a pixel-to-pixel manner. The proposed method performs well in stain normalization and achieves a better accuracy and image quality.