Search

Efficient point cloud compression is essential for applications like virtual and mixed reality, autonomous driving, and cultural heritage. This paper proposes a deep learning-based inter-frame encoding scheme for dynamic point cloud geometry compression. We propose a lossy geometry compression scheme that predicts the latent representation of the current frame using the previous frame by employing a novel feature space inter-prediction network. The proposed network utilizes sparse convolutions with hierarchical multiscale 3D feature learning to encode the current frame using the previous frame. The proposed method introduces a novel predictor network for motion compensation in the feature domain to map the latent representation of the previous frame to the coordinates of the current frame to predict the current frame’s feature embedding. The framework transmits the residual of the predicted features and the actual features by compressing them using a learned probabilistic factorized entropy model. At the receiver, the decoder hierarchically reconstructs the current frame by progressively rescaling the feature embedding. The proposed framework is compared to the state-of-the-art Video-based Point Cloud Compression (V-PCC) and Geometry-based Point Cloud Compression (G-PCC) schemes standardized by the Moving Picture Experts Group (MPEG). The proposed method achieves more than 88% BD-Rate (Bjøntegaard Delta Rate) reduction against G-PCCv20 Octree, more than 56% BD-Rate savings against G-PCCv20 Trisoup, more than 62% BD-Rate reduction against V-PCC intra-frame encoding mode, and more than 52% BD-Rate savings against V-PCC P-frame-based inter-frame encoding mode using HEVC. These significant performance gains are cross-checked and verified in the MPEG working group. [ABSTRACT FROM AUTHOR]

Recent developments in the field of non-local attention (NLA) have led to a renewed interest in self-similarity-based single image super-resolution (SISR). Researchers usually use the NLA to explore non-local self-similarity (NSS) in SISR and achieve satisfactory reconstruction results. However, a surprising phenomenon that the reconstruction performance of the standard NLA is similar to that of the NLA with randomly selected regions prompted us to revisit NLA. In this paper, we first analyzed the attention map of the standard NLA from different perspectives and discovered that the resulting probability distribution always has full support for every local feature, which implies a statistical waste of assigning values to irrelevant non-local features, especially for SISR which needs to model long-range dependence with a large number of redundant non-local features. Based on these findings, we introduced a concise yet effective soft thresholding operation to obtain high-similarity-pass attention (HSPA), which is beneficial for generating a more compact and interpretable distribution. Furthermore, we derived some key properties of the soft thresholding operation that enable training our HSPA in an end-to-end manner. The HSPA can be integrated into existing deep SISR models as an efficient general building block. In addition, to demonstrate the effectiveness of the HSPA, we constructed a deep high-similarity-pass attention network (HSPAN) by integrating a few HSPAs in a simple backbone. Extensive experimental results demonstrate that HSPAN outperforms state-of-the-art approaches on both quantitative and qualitative evaluations. Our code and a pre-trained model were uploaded to GitHub (https://github.com/laoyangui/HSPAN) for validation. [ABSTRACT FROM AUTHOR]

The vanilla Few-shot Learning (FSL) learns to build a classifier for a new concept from one or very few target examples, with the general assumption that source and target classes are sampled from the same domain. Recently, the task of Cross-Domain Few-Shot Learning (CD-FSL) aims at tackling the FSL where there is a huge domain shift between the source and target datasets. Extensive efforts on CD-FSL have been made via either directly extending the meta-learning paradigm of vanilla FSL methods, or employing massive unlabeled target data to help learn models. In this paper, we notice that in the CD-FSL task, the few labeled target images have never been explicitly leveraged to inform the model in the training stage. However, such a labeled target example set is very important to bridge the huge domain gap. Critically, this paper advocates a more practical training scenario for CD-FSL. And our key insight is to utilize a few labeled target data to guide the learning of the CD-FSL model. Technically, we propose a novel Generalized Meta-learning based Feature-Disentangled Mixup network, namely GMeta-FDMixup. We make three key contributions of utilizing GMeta-FDMixup to address CD-FSL. Firstly, we present two mixup modules – mixup-P and mixup-M that help facilitate utilizing the unbalanced and disjoint source and target datasets. These two novel modules enable diverse image generation for training the model on the source domain. Secondly, to narrow the domain gap explicitly, we contribute a novel feature disentanglement module that learns to decouple the domain-irrelevant and domain-specific features. By stripping the domain-specific features, we alleviate the negative effects caused by the domain inductive bias. Finally, we repurpose a new contrastive learning module, dubbed ConL. ConL prevents the model from only capturing category-related features via introducing contrastive loss. Thus, the generalization ability on novel categories is improved. Extensive experimental results on two benchmarks show the superiority of our setting and the effectiveness of our method. Code and models will be released. [ABSTRACT FROM AUTHOR]

Recent research advances in salient object detection (SOD) could largely be attributed to ever-stronger multi-scale feature representation empowered by the deep learning technologies. The existing SOD deep models extract multi-scale features via the off-the-shelf encoders and combine them smartly via various delicate decoders. However, the kernel sizes in this commonly-used thread are usually “fixed”. In our new experiments, we have observed that kernels of small size are preferable in scenarios containing tiny salient objects. In contrast, large kernel sizes could perform better for images with large salient objects. Inspired by this observation, we advocate the “dynamic” scale routing (as a brand-new idea) in this paper. It will result in a generic plug-in that could directly fit the existing feature backbone. This paper’s key technical innovations are two-fold. First, instead of using the vanilla convolution with fixed kernel sizes for the encoder design, we propose the dynamic pyramid convolution (DPConv), which dynamically selects the best-suited kernel sizes w.r.t. the given input. Second, we provide a self-adaptive bidirectional decoder design to accommodate the DPConv-based encoder best. The most significant highlight is its capability of routing between feature scales and their dynamic collection, making the inference process scale-aware. As a result, this paper continues to enhance the current SOTA performance. Both the code and dataset are publicly available at https://github.com/wuzhenyubuaa/DPNet. [ABSTRACT FROM AUTHOR]

Learning-based infrared small object detection methods currently rely heavily on the classification backbone network. This tends to result in tiny object loss and feature distinguishability limitations as the network depth increases. Furthermore, small objects in infrared images are frequently emerged bright and dark, posing severe demands for obtaining precise object contrast information. For this reason, we in this paper propose a simple and effective “U-Net in U-Net” framework, UIU-Net for short, and detect small objects in infrared images. As the name suggests, UIU-Net embeds a tiny U-Net into a larger U-Net backbone, enabling the multi-level and multi-scale representation learning of objects. Moreover, UIU-Net can be trained from scratch, and the learned features can enhance global and local contrast information effectively. More specifically, the UIU-Net model is divided into two modules: the resolution-maintenance deep supervision (RM-DS) module and the interactive-cross attention (IC-A) module. RM-DS integrates Residual U-blocks into a deep supervision network to generate deep multi-scale resolution-maintenance features while learning global context information. Further, IC-A encodes the local context information between the low-level details and high-level semantic features. Extensive experiments conducted on two infrared single-frame image datasets, i.e., SIRST and Synthetic datasets, show the effectiveness and superiority of the proposed UIU-Net in comparison with several state-of-the-art infrared small object detection methods. The proposed UIU-Net also produces powerful generalization performance for video sequence infrared small object datasets, e.g., ATR ground/air video sequence dataset. The codes of this work are available openly at https://github.com/danfenghong/IEEE [ABSTRACT FROM AUTHOR]

Conventional stereoscopic displays suffer from vergence-accommodation conflict and cause visual fatigue. Integral-imaging-based displays resolve the problem by directly projecting the sub-aperture views of a light field into the eyes using a microlens array or a similar structure. However, such displays have an inherent trade-off between angular and spatial resolutions. In this paper, we propose a novel coded time-division multiplexing technique that projects encoded sub-aperture views to the eyes of a viewer with correct cues for vergence-accommodation reflex. Given sparse light field sub-aperture views, our pipeline can provide a perception of high-resolution refocused images with minimal aliasing by jointly optimizing the sub-aperture views for display and the coded aperture pattern. This is achieved via deep learning in an end-to-end fashion by simulating light transport and image formation with Fourier optics. To our knowledge, this work is among the first that optimize the light field display pipeline with deep learning. We verify our idea with objective image quality metrics (PSNR, SSIM, and LPIPS) and perform an extensive study on various customizable design variables in our display pipeline. Experimental results show that light fields displayed using the proposed technique indeed have higher quality than that of baseline display designs. [ABSTRACT FROM AUTHOR]

Scene flow represents the 3D motion of every point in the dynamic environments. Like the optical flow that represents the motion of pixels in 2D images, 3D motion representation of scene flow benefits many applications, such as autonomous driving and service robot. This paper studies the problem of scene flow estimation from two consecutive 3D point clouds. In this paper, a novel hierarchical neural network with double attention is proposed for learning the correlation of point features in adjacent frames and refining scene flow from coarse to fine layer by layer. The proposed network has a new more-for-less hierarchical architecture. The more-for-less means that the number of input points is greater than the number of output points for scene flow estimation, which brings more input information and balances the precision and resource consumption. In this hierarchical architecture, scene flow of different levels is generated and supervised respectively. A novel attentive embedding module is introduced to aggregate the features of adjacent points using a double attention method in a patch-to-patch manner. The proper layers for flow embedding and flow supervision are carefully considered in our network designment. Experiments show that the proposed network outperforms the state-of-the-art performance of 3D scene flow estimation on the FlyingThings3D and KITTI Scene Flow 2015 datasets. We also apply the proposed network to the realistic LiDAR odometry task, which is a key problem in autonomous driving. The experiment results demonstrate that our proposed network can outperform the ICP-based method and shows good practical application ability. The source codes will be released on https://github.com/IRMVLab/HALFlow. [ABSTRACT FROM AUTHOR]

Face hallucination or super-resolution is a practical application of general image super-resolution which has been recently studied by many researchers. The challenge of good face hallucination comes from a variety of poses, illuminations, facial expressions, and other degradations. In many proposed methods, researchers resolve it by using a generative neural network to reduce the perceptual loss so we can generate a photo-realistic image. The problem is that researchers usually overlook the fidelity of the super-resolved image which could affect further facial image processing. Meanwhile, many CNN based approaches cascade multiple networks to extract facial prior information to improve super-resolution quality. Because of the end-to-end design, the details are missing for investigation. In this paper, we combine new techniques in convolutional neural network and random forests to a Hierarchical CNN based Random Forests (HCRF) approach for face super-resolution in a coarse-to-fine manner. In the proposed approach, we focus on a general approach that can handle facial images with various conditions without pre-processing. To the best of our knowledge, this is the first paper that combines the advantages of deep learning with random forests for face super-resolution. To achieve superior performance, we propose two novel CNN models for coarse facial image super-resolution and segmentation and then apply new random forests to target on local facial features refinement making use of the segmentation results. Extensive benchmark experiments on subjective and objective evaluation show that HCRF can achieve comparable speed and competitive performance compared with state-of-the-art super-resolution approaches for very low-resolution images. [ABSTRACT FROM AUTHOR]

This paper focuses on the thorax disease classification problem in chest X-ray (CXR) images. Different from the generic image classification task, a robust and stable CXR image analysis system should consider the unique characteristics of CXR images. Particularly, it should be able to: 1) automatically focus on the disease-critical regions, which usually are of small sizes; 2) adaptively capture the intrinsic relationships among different disease features and utilize them to boost the multi-label disease recognition rates jointly. In this paper, we propose to learn discriminative features with a two-branch architecture, named ConsultNet, to achieve those two purposes simultaneously. ConsultNet consists of two components. First, an information bottleneck constrained feature selector extracts critical disease-specific features according to the feature importance. Second, a spatial-and-channel encoding based feature integrator enhances the latent semantic dependencies in the feature space. ConsultNet fuses these discriminative features to improve the performance of thorax disease classification in CXRs. Experiments conducted on the ChestX-ray14 and CheXpert dataset demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Light field imaging, which captures both spatial and angular information, improves user immersion by enabling post-capture actions, such as refocusing and changing view perspective. However, light fields represent very large volumes of data with a lot of redundancy that coding methods try to remove. State-of-the-art coding methods indeed usually focus on improving compression efficiency and overlook other important features in light field compression such as scalability. In this paper, we propose a novel light field image compression method that enables (i) viewport scalability, (ii) quality scalability, (iii) spatial scalability, (iv) random access, and (v) uniform quality distribution among viewports, while keeping compression efficiency high. To this end, light fields in each spatial resolution are divided into sequential viewport layers, and viewports in each layer are encoded using the previously encoded viewports. In each viewport layer, the available viewports are used to synthesize intermediate viewports using a video interpolation deep learning network. The synthesized views are used as virtual reference images to enhance the quality of intermediate views. An image super-resolution method is applied to improve the quality of the lower spatial resolution layer. The super-resolved images are also used as virtual reference images to improve the quality of the higher spatial resolution layer. The proposed structure also improves the flexibility of light field streaming, provides random access to the viewports, and increases error resiliency. The experimental results demonstrate that the proposed method achieves a high compression efficiency and it can adapt to the display type, transmission channel, network condition, processing power, and user needs. [ABSTRACT FROM AUTHOR]

Semantic segmentation methods based on deep neural networks have achieved great success in recent years. However, training such deep neural networks relies heavily on a large number of images with accurate pixel-level labels, which requires a huge amount of human effort, especially for large-scale remote sensing images. In this paper, we propose a point-based weakly supervised learning framework called the deep bilateral filtering network (DBFNet) for the semantic segmentation of remote sensing images. Compared with pixel-level labels, point annotations are usually sparse and cannot reveal the complete structure of the objects; they also lack boundary information, thus resulting in incomplete prediction within the object and the loss of object boundaries. To address these problems, we incorporate the bilateral filtering technique into deeply learned representations in two respects. First, since a target object contains smooth regions that always belong to the same category, we perform deep bilateral filtering (DBF) to filter the deep features by a nonlinear combination of nearby feature values, which encourages the nearby and similar features to become closer, thus achieving a consistent prediction in the smooth region. In addition, the DBF can distinguish the boundary by enlarging the distance between the features on different sides of the edge, thus preserving the boundary information well. Experimental results on two widely used datasets, the ISPRS 2-D semantic labeling Potsdam and Vaihingen datasets, demonstrate that our proposed DBFNet can achieve a highly competitive performance compared with state-of-the-art fully-supervised methods. Code is available at https://github.com/Luffy03/DBFNet. [ABSTRACT FROM AUTHOR]

Hyperspectral videos can provide the spatial, spectral, and motion information of targets, which makes it possible to track camouflaged targets that are similar to the background. However, hyperspectral object tracking is a challenging task, due to the huge hyperspectral video data dimension and the “data hungry” problem for the model training. Insufficient training data can seriously interfere with the accuracy and generalization of the tracking models. In this paper, a dual deep Siamese network framework for hyperspectral object tracking (SiamHYPER) is proposed for learning a hyperspectral tracker from a pretrained RGB tracker in the case of the “data hungry” problem. Specifically, in addition to a pretrained RGB-based Siamese tracker, a hyperspectral target-aware module is designed to mine the spectral information during the target prediction, and a spatial-spectral cross-attention module is introduced to further fuse the deep spatial and spectral features extracted from the RGB tracker and the hyperspectral target-aware module. Benefiting from the guidance training of the RGB tracker, a robust hyperspectral object tracker can be trained effectively with only a small number of hyperspectral video samples, to overcome the “data hungry” problem. In the experiments conducted in this study, the SiamHYPER framework was verified using SiamBAN and SiamRPN++, with 13 000 frames of hyperspectral videos for training, and achieved the best performance on the publicly available hyperspectral dataset released as part of the WHISPERS Hyperspectral Object Tracking Challenge. The area under the curve (AUC) of SiamHYPER was increased by nearly 8.9% and 7.2%, respectively, when compared with the current state-of-the-art RGB-based and hyperspectral trackers. In addition, the processing speed of SiamHYPER was 19 FPS, which is much higher than that of the current state-of-the-art hyperspectral trackers. The source code is available at zhenliuzhenqi/HOT: Hyperspectral object tracking (github.com). [ABSTRACT FROM AUTHOR]

Acoustic images are an emergent data modality for multimodal scene understanding. Such images have the peculiarity of distinguishing the spectral signature of the sound coming from different directions in space, thus providing a richer information as compared to that derived from single or binaural microphones. However, acoustic images are typically generated by cumbersome and costly microphone arrays which are not as widespread as ordinary microphones. This paper shows that it is still possible to generate acoustic images from off-the-shelf cameras equipped with only a single microphone and how they can be exploited for audio-visual scene understanding. We propose three architectures inspired by Variational Autoencoder, U-Net and adversarial models, and we assess their advantages and drawbacks. Such models are trained to generate spatialized audio by conditioning them to the associated video sequence and its corresponding monaural audio track. Our models are trained using the data collected by a microphone array as ground truth. Thus they learn to mimic the output of an array of microphones in the very same conditions. We assess the quality of the generated acoustic images considering standard generation metrics and different downstream tasks (classification, cross-modal retrieval and sound localization). We also evaluate our proposed models by considering multimodal datasets containing acoustic images, as well as datasets containing just monaural audio signals and RGB video frames. In all of the addressed downstream tasks we obtain notable performances using the generated acoustic data, when compared to the state of the art and to the results obtained using real acoustic images as input. [ABSTRACT FROM AUTHOR]

Modern top-performing object detectors depend heavily on backbone networks, whose advances bring consistent performance gains through exploring more effective network structures. In this paper, we propose a novel and flexible backbone framework, namely CBNet, to construct high-performance detectors using existing open-source pre-trained backbones under the pre-training fine-tuning paradigm. In particular, CBNet architecture groups multiple identical backbones, which are connected through composite connections. Specifically, it integrates the high- and low-level features of multiple identical backbone networks and gradually expands the receptive field to more effectively perform object detection. We also propose a better training strategy with auxiliary supervision for CBNet-based detectors. CBNet has strong generalization capabilities for different backbones and head designs of the detector architecture. Without additional pre-training of the composite backbone, CBNet can be adapted to various backbones (i.e., CNN-based vs. Transformer-based) and head designs of most mainstream detectors (i.e., one-stage vs. two-stage, anchor-based vs. anchor-free-based). Experiments provide strong evidence that, compared with simply increasing the depth and width of the network, CBNet introduces a more efficient, effective, and resource-friendly way to build high-performance backbone networks. Particularly, our CB-Swin-L achieves 59.4% box AP and 51.6% mask AP on COCO test-dev under the single-model and single-scale testing protocol, which are significantly better than the state-of-the-art results (i.e., 57.7% box AP and 50.2% mask AP) achieved by Swin-L, while reducing the training time by $6\times $. With multi-scale testing, we push the current best single model result to a new record of 60.1% box AP and 52.3% mask AP without using extra training data. Code is available at https://github.com/VDIGPKU/CBNetV2 . [ABSTRACT FROM AUTHOR]

Recently, deep learning based multispectral (MS) and panchromatic (PAN) image fusion methods have been proposed, which extracted features automatically and hierarchically by a series of non-linear transformations to model the complicated imaging discrepancy. But they always pay more attention to the extraction and compensation of spatial details and use the mean squared error or mean absolute error as a loss function, regardless of the preservation of spectral information contained in multispectral images. For the sake of the improvements in both spatial and spectral resolution, this paper presents a novel fusion model that takes the spectral preservation into consideration, and learns the spectral cues from the process of generating a spectrally refined multispectral image, which is constrained by a spectral loss between the generated image and the reference image. Then these spectral cues are used to modulate the PAN features to obtain final fusion result. Experimental results on reduced-resolution and full-resolution datasets demonstrate that the proposed method can obtain a better fusion result in terms of visual inspection and evaluation indices when compared with current state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Visible-infrared person re-identification (VI-ReID) task aims to retrieve the same person between visible and infrared images. VI-ReID is challenging as the images captured by different spectra present large cross-modality discrepancy. Many methods adopt a two-stream network and design additional constraint conditions to extract shared features for different modalities. However, the interaction between the feature extraction processes of different modalities is rarely considered. In this paper, a partially interactive collaboration method is proposed to exploit the complementary information of different modalities to reduce the modality gap for VI-ReID. Specifically, the proposed method is achieved in a partially interactive-shared architecture: collaborative shallow layers and shared deep layers. The collaborative shallow layers consider the interaction between modality-specific features of different modalities, encouraging the feature extraction processes of different modalities constrain each other to enhance feature representations. The shared deep layers further embed the modality-specific features to a common space to endow them the same identity discriminability. To ensure the interactive collaborative learning implement effectively, the conventional loss and collaborative loss are utilized jointly to train the whole network. Extensive experiments on two publicly available VI-ReID datasets verify the superiority of the proposed PIC method. Specifically, the proposed method achieves a rank-1 accuracy of 83.6% and 57.5% on RegDB and SYSU-MM01 datasets, respectively. [ABSTRACT FROM AUTHOR]

Well-known compressed sensing (CS) is widely used in image acquisition and reconstruction. However, accurately reconstructing images from measurements at low sampling rates remains a considerable challenge. In this paper, we propose a novel Transformer-based hybrid architecture (dubbed TransCS) to achieve high-quality image CS. In the sampling module, TransCS adopts a trainable sensing matrix strategy that gains better image reconstruction by learning the structural information from the training images. In the reconstruction module, inspired by the powerful long-distance dependence modelling capacity of the Transformer, a customized iterative shrinkage-thresholding algorithm (ISTA)-based Transformer backbone that iteratively works with gradient descent and soft threshold operation is designed to model the global dependency among image subblocks. Moreover, the auxiliary convolutional neural network (CNN) is introduced to capture the local features of images. Therefore, the proposed hybrid architecture that integrates the customized ISTA-based Transformer backbone with CNN can gain high-performance reconstruction for image compressed sensing. The experimental results demonstrate that our proposed TransCS obtains superior reconstruction quality and noise robustness on several public benchmark datasets compared with other state-of-the-art methods. Our code is available on TransCS. [ABSTRACT FROM AUTHOR]

High-precision road detection from very high resolution (VHR) remote sensing images has broad application value. However, the most advanced deep learning based methods often fail to identify roads when there is a distribution discrepancy between the training samples and test samples, due to their limited generalization ability. In this paper, to address this problem, an open-source data-driven domain-specific representation (OSM-DOER) framework is proposed for cross-domain road detection. On the one hand, as the spatial structure information of the source and target domains is similar, but the texture information is different, the domain-specific representation (DOER) framework is proposed, which not only aligns the distributions of the spatial structure information, but also learns the domain-specific texture information. Furthermore, in order to enhance the representation of the target domain data distribution, open-source and freely available OpenStreetMap (OSM) road centerline data are utilized to generate target domain samples, which are then used in the network training as the supervised information for the target domain. Finally, to verify the superiority of the proposed OSM-DOER framework, we conducted extensive experiments with the public SpaceNet and DeepGlobe road datasets, and large-scale road datasets from Birmingham in the UK and Shanghai in China. The experimental results demonstrate that the proposed OSM-DOER framework shows obvious advantages over the mainstream road detection methods, and the use of OSM road centerline data has great potential for the road detection task. [ABSTRACT FROM AUTHOR]

Due to the wavelength-dependent light attenuation, refraction and scattering, underwater images usually suffer from color distortion and blurred details. However, due to the limited number of paired underwater images with undistorted images as reference, training deep enhancement models for diverse degradation types is quite difficult. To boost the performance of data-driven approaches, it is essential to establish more effective learning mechanisms that mine richer supervised information from limited training sample resources. In this paper, we propose a novel underwater image enhancement network, called SGUIE-Net, in which we introduce semantic information as high-level guidance via region-wise enhancement feature learning. Accordingly, we propose semantic region-wise enhancement module to better learn local enhancement features for semantic regions with multi-scale perception. After using them as complementary features and feeding them to the main branch, which extracts the global enhancement features on the original image scale, the fused features bring semantically consistent and visually superior enhancements. Extensive experiments on the publicly available datasets and our proposed dataset demonstrate the impressive performance of SGUIE-Net. The code and proposed dataset are available at https://trentqq.github.io/SGUIE-Net.html . [ABSTRACT FROM AUTHOR]

Generation of a 3D model of an object from multiple views has a wide range of applications. Different parts of an object would be accurately captured by a particular view or a subset of views in the case of multiple views. In this paper, a novel coarse-to-fine network (C2FNet) is proposed for 3D point cloud generation from multiple views. C2FNet generates subsets of 3D points that are best captured by individual views with the support of other views in a coarse-to-fine way, and then fuses these subsets of 3D points to a whole point cloud. It consists of a coarse generation module where coarse point clouds are constructed from multiple views by exploring the cross-view spatial relations, and a fine generation module where the coarse point cloud features are refined under the guidance of global consistency in appearance and context. Extensive experiments on the benchmark datasets have demonstrated that the proposed method outperforms the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

State-of-the-art 2D image compression schemes rely on the power of convolutional neural networks (CNNs). Although CNNs offer promising perspectives for 2D image compression, extending such models to omnidirectional images is not straightforward. First, omnidirectional images have specific spatial and statistical properties that can not be fully captured by current CNN models. Second, basic mathematical operations composing a CNN architecture, e.g., translation and sampling, are not well-defined on the sphere. In this paper, we study the learning of representation models for omnidirectional images and propose to use the properties of HEALPix uniform sampling of the sphere to redefine the mathematical tools used in deep learning models for omnidirectional images. In particular, we: i) propose the definition of a new convolution operation on the sphere that keeps the high expressiveness and the low complexity of a classical 2D convolution; ii) adapt standard CNN techniques such as stride, iterative aggregation, and pixel shuffling to the spherical domain; and then iii) apply our new framework to the task of omnidirectional image compression. Our experiments show that our proposed on-the-sphere solution leads to a better compression gain that can save 13.7% of the bit rate compared to similar learned models applied to equirectangular images. Also, compared to learning models based on graph convolutional networks, our solution supports more expressive filters that can preserve high frequencies and provide a better perceptual quality of the compressed images. Such results demonstrate the efficiency of the proposed framework, which opens new research venues for other omnidirectional vision tasks to be effectively implemented on the sphere manifold. [ABSTRACT FROM AUTHOR]

Most existing RGB-D salient object detection (SOD) methods are primarily focusing on cross-modal and cross-level saliency fusion, which has been proved to be efficient and effective. However, these methods still have a critical limitation, i.e., their fusion patterns – typically the combination of selective characteristics and its variations, are too highly dependent on the network’s non-linear adaptability. In such methods, the balances between RGB and D (Depth) are formulated individually considering the intermediate feature slices, but the relation at the modality level may not be learned properly. The optimal RGB-D combinations differ depending on the RGB-D scenarios, and the exact complementary status is frequently determined by multiple modality-level factors, such as D quality, the complexity of the RGB scene, and degree of harmony between them. Therefore, given the existing approaches, it may be difficult for them to achieve further performance breakthroughs, as their methodologies belong to some methods that are somewhat less modality sensitive. To conquer this problem, this paper presents the Modality-aware Decoder (MaD). The critical technical innovations include a series of feature embedding, modality reasoning, and feature back-projecting and collecting strategies, all of which upgrade the widely-used multi-scale and multi-level decoding process to be modality-aware. Our MaD achieves competitive performance over other state-of-the-art (SOTA) models without using any fancy tricks in the decoder’s design. Codes and results will be publicly available at https://github.com/MengkeSong/MaD. [ABSTRACT FROM AUTHOR]

Lighting is a determining factor in photography that affects the style, expression of emotion, and even quality of images. Creating or finding satisfying lighting conditions, in reality, is laborious and time-consuming, so it is of great value to develop a technology to manipulate illumination in an image as post-processing. Although previous works have explored techniques based on the physical viewpoint for relighting images, extensive supervisions and prior knowledge are necessary to generate reasonable images, restricting the generalization ability of these works. In contrast, we take the viewpoint of image-to-image translation and implicitly merge ideas of the conventional physical viewpoint. In this paper, we present an Illumination-Aware Network (IAN) which follows the guidance from hierarchical sampling to progressively relight a scene from a single image with high efficiency. In addition, an Illumination-Aware Residual Block (IARB) is designed to approximate the physical rendering process and to extract precise descriptors of light sources for further manipulations. We also introduce a depth-guided geometry encoder for acquiring valuable geometry- and structure-related representations once the depth information is available. Experimental results show that our proposed method produces better quantitative and qualitative relighting results than previous state-of-the-art methods. The code and models are publicly available on https://github.com/NK-CS-ZZL/IAN. [ABSTRACT FROM AUTHOR]

Deep learning (DL) based hyperspectral images (HSIs) denoising approaches directly learn the nonlinear mapping between noisy and clean HSI pairs. They usually do not consider the physical characteristics of HSIs. This drawback makes the models lack interpretability that is key to understanding their denoising mechanism and limits their denoising ability. In this paper, we introduce a novel model-guided interpretable network for HSI denoising to tackle this problem. Fully considering the spatial redundancy, spectral low-rankness, and spectral-spatial correlations of HSIs, we first establish a subspace-based multidimensional sparse (SMDS) model under the umbrella of tensor notation. After that, the model is unfolded into an end-to-end network named SMDS-Net, whose fundamental modules are seamlessly connected with the denoising procedure and optimization of the SMDS model. This makes SMDS-Net convey clear physical meanings, i.e., learning the low-rankness and sparsity of HSIs. Finally, all key variables are obtained by discriminative training. Extensive experiments and comprehensive analysis on synthetic and real-world HSIs confirm the strong denoising ability, strong learning capability, promising generalization ability, and high interpretability of SMDS-Net against the state-of-the-art HSI denoising methods. The source code and data of this article will be made publicly available at https://github.com/bearshng/smds-net for reproducible research. [ABSTRACT FROM AUTHOR]

Domain adaptation is to transfer the shared knowledge learned from the source domain to a new environment, i.e., target domain. One common practice is to train the model on both labeled source-domain data and unlabeled target-domain data. Yet the learned models are usually biased due to the strong supervision of the source domain. Most researchers adopt the early-stopping strategy to prevent over-fitting, but when to stop training remains a challenging problem since the lack of the target-domain validation set. In this paper, we propose one efficient bootstrapping method, called Adaboost Student, explicitly learning complementary models during training and liberating users from empirical early stopping. Adaboost Student combines deep model learning with the conventional training strategy, i.e., adaptive boosting, and enables interactions between learned models and the data sampler. We adopt one adaptive data sampler to progressively facilitate learning on hard samples and aggregate “weak” models to prevent over-fitting. Extensive experiments show that (1) Without the need to worry about the stopping time, AdaBoost Student provides one robust solution by efficient complementary model learning during training. (2) AdaBoost Student is orthogonal to most domain adaptation methods, which can be combined with existing approaches to further improve the state-of-the-art performance. We have achieved competitive results on three widely-used scene segmentation domain adaptation benchmarks. [ABSTRACT FROM AUTHOR]

Recently, embedding and metric-based few-shot learning (FSL) has been introduced into hyperspectral image classification (HSIC) and achieved impressive progress. To further enhance the performance with few labeled samples, we in this paper propose a novel FSL framework for HSIC with a class-covariance metric (CMFSL). Overall, the CMFSL learns global class representations for each training episode by interactively using training samples from the base and novel classes, and a synthesis strategy is employed on the novel classes to avoid overfitting. During the meta-training and meta-testing, the class labels are determined directly using the Mahalanobis distance measurement rather than an extra classifier. Benefiting from the task-adapted class-covariance estimations, the CMFSL can construct more flexible decision boundaries than the commonly used Euclidean metric. Additionally, a lightweight cross-scale convolutional network (LXConvNet) consisting of 3D and 2D convolutions is designed to thoroughly exploit the spectral-spatial information in the high-frequency and low-frequency scales with low computational complexity. Furthermore, we devise a spectral-prior-based refinement module (SPRM) in the initial stage of feature extraction, which cannot only force the network to emphasize the most informative bands while suppressing the useless ones, but also alleviate the effects of the domain shift between the base and novel categories to learn a collaborative embedding mapping. Extensive experiment results on four benchmark data sets demonstrate that the proposed CMFSL can outperform the state-of-the-art methods with few-shot annotated samples. [ABSTRACT FROM AUTHOR]

Deep learning algorithms have obtained great success in semantic segmentation of very high-resolution (VHR) remote sensing images. Nevertheless, training these models generally requires a large amount of accurate pixel-wise annotations, which is very laborious and time-consuming to collect. To reduce the annotation burden, this paper proposes a consistency-regularized region-growing network (CRGNet) to achieve semantic segmentation of VHR remote sensing images with point-level annotations. The key idea of CRGNet is to iteratively select unlabeled pixels with high confidence to expand the annotated area from the original sparse points. However, since there may exist some errors and noises in the expanded annotations, directly learning from them may mislead the training of the network. To this end, we further propose the consistency regularization strategy, where a base classifier and an expanded classifier are employed. Specifically, the base classifier is supervised by the original sparse annotations, while the expanded classifier aims to learn from the expanded annotations generated by the base classifier with the region-growing mechanism. The consistency regularization is thereby achieved by minimizing the discrepancy between the predictions from both the base and the expanded classifiers. We find such a simple regularization strategy is yet very useful to control the quality of the region-growing mechanism. Extensive experiments on two benchmark datasets demonstrate that the proposed CRGNet significantly outperforms the existing state-of-the-art methods. Codes and pre-trained models are available online (https://github.com/YonghaoXu/CRGNet). [ABSTRACT FROM AUTHOR]

Reducing redundancy is crucial for improving the efficiency of video recognition models. An effective approach is to select informative content from the holistic video, yielding a popular family of dynamic video recognition methods. However, existing dynamic methods focus on either temporal or spatial selection independently while neglecting a reality that the redundancies are usually spatial and temporal, simultaneously. Moreover, their selected content is usually cropped with fixed shapes (e.g., temporally-cropped frames, spatially-cropped patches), while the realistic distribution of informative content can be much more diverse. With these two insights, this paper proposes to integrate temporal and spatial selection into an Action Keypoint Network (AK-Net). From different frames and positions, AK-Net selects some informative points scattered in arbitrary-shaped regions as a set of “action keypoints” and then transforms the video recognition into point cloud classification. More concretely, AK-Net has two steps, i.e., the keypoint selection and the point cloud classification. First, it inputs the video into a baseline network and outputs a feature map from an intermediate layer. We view each pixel on this feature map as a spatial-temporal point and select some informative keypoints using self-attention. Second, AK-Net devises a ranking criterion to arrange the keypoints into an ordered 1D sequence. Since the video is represented with a 1D sequence after the specified layer, AK-Net transforms the subsequent layers into a point cloud classification sub-net by compacting the original 2D convolutional kernels into 1D kernels. Consequentially, AK-Net brings two-fold benefits for efficiency: The keypoint selection step collects informative content within arbitrary shapes and increases the efficiency for modeling spatial-temporal dependencies, while the point cloud classification step further reduces the computational cost by compacting the convolutional kernels. Experimental results show that AK-Net can consistently improve the efficiency and performance of baseline methods on several video recognition benchmarks. [ABSTRACT FROM AUTHOR]

Deep subspace learning is an important branch of self-supervised learning and has been a hot research topic in recent years, but current methods do not fully consider the individualities of temporal data and related tasks. In this paper, by transforming the individualities of motion capture data and segmentation task as the supervision, we propose the local self-expression subspace learning network. Specifically, considering the temporality of motion data, we use the temporal convolution module to extract temporal features. To implement the local validity of self-expression in temporal tasks, we design the local self-expression layer which only maintains the representation relations with temporally adjacent motion frames. To simulate the interpolatability of motion data in the feature space, we impose a group sparseness constraint on the local self-expression layer to impel the representations only using selected keyframes. Besides, based on the subspace assumption, we propose the subspace projection loss, which is induced from distances of each frame projected to the fitted subspaces, to penalize the potential clustering errors. The superior performances of the proposed model on the segmentation task of synthetic data and three tasks of real motion capture data demonstrate the feature learning ability of our model. [ABSTRACT FROM AUTHOR]

Serving as an essential step for many applications of image processing, superpixel generation has attracted a lot of attentions. Most existing superpixel generation algorithms focus on the boundary adherence and compactness of the superpixels, but ignore the topological consistency between the superpixels, which severely limites their applications in the subsequent tasks, especially in the CNN based image processing tasks. In this paper, we present a fast lattice superpixel generation algorithm, which can generate superpixels with lattice topology like the original pixels. We also propose a local similarity loss function to improve the segmentation accuracy of the generated lattice superpixels. The whole algorithm is parallelly implemented on GPU. We perform extensive experiments on three datasets (i.e., BSDS500, NYUv2 and VOC) to verify the efficacy of our algorithm. The experimental results show that our method achieves competitive results compared to the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel multi-scale attention based network (called MSA-Net) for feature matching problems. Current deep networks based feature matching methods suffer from limited effectiveness and robustness when applied to different scenarios, due to random distributions of outliers and insufficient information learning. To address this issue, we propose a multi-scale attention block to enhance the robustness to outliers, for improving the representational ability of the feature map. In addition, we also design a novel context channel refine block and a context spatial refine block to mine the information context with less parameters along channel and spatial dimensions, respectively. The proposed MSA-Net is able to effectively infer the probability of correspondences being inliers with less parameters. Extensive experiments on outlier removal and relative pose estimation have shown the performance improvements of our network over current state-of-the-art methods with less parameters on both outdoor and indoor datasets. Notably, our proposed network achieves an 11.7% improvement at error threshold 5° without RANSAC than the state-of-the-art method on relative pose estimation task when trained on YFCC100M dataset. [ABSTRACT FROM AUTHOR]

Previous deep learning-based video stabilizers require a large scale of paired unstable and stable videos for training, which are difficult to collect. Traditional trajectory-based stabilizers, on the other hand, divide the task into several sub-tasks and tackle them subsequently, which are fragile in textureless and occluded regions regarding the usage of hand-crafted features. In this paper, we attempt to tackle the video stabilization problem in a deep unsupervised learning manner, which borrows the divide-and-conquer idea from traditional stabilizers while leveraging the representation power of DNNs to handle the challenges in real-world scenarios. Technically, DUT is composed of a trajectory estimation stage and a trajectory smoothing stage. In the trajectory estimation stage, we first estimate the motion of keypoints, initialize and refine the motion of grids via a novel multi-homography estimation strategy and a motion refinement network, respectively, and get the grid-based trajectories via temporal association. In the trajectory smoothing stage, we devise a novel network to predict dynamic smoothing kernels for trajectory smoothing, which can well adapt to trajectories with different dynamic patterns. We exploit the spatial and temporal coherence of keypoints and grid vertices to formulate the training objectives, resulting in an unsupervised training scheme. Experiment results on public benchmarks show that DUT outperforms state-of-the-art methods both qualitatively and quantitatively. The source code is available at https://github.com/Annbless/DUTCode. [ABSTRACT FROM AUTHOR]

Multi-focus image fusion (MFIF) attempts to achieve an “all-focused” image from multiple source images with the same scene but different focused objects. Given the lack of multi-focus image sets for network training, we propose a self-supervised residual feature learning model in this paper. The model consists of a feature extraction network and a fusion module. We select image super-resolution as a pretext task in the MFIF field, which is supported by a new residual gradient prior discovered by our theoretical study for low- and high-resolution (LR-HR) image pairs, as well as for multi-focus images. In the pretext task, our network’s training set is LR-HR image pairs generated from natural images, and HR images can be regarded as pseudo-labels of LR images. In the fusion task, the trained network extracts residual features of multi-focus images firstly. Secondly, the fusion module, consisting of an activity level measurement and a new boundary refinement method, is leveraged for the features to generated decision maps. Experimental results, both subjective evaluations and objective evaluations, demonstrate that our approach outperforms other state-of-the-art fusion algorithms. [ABSTRACT FROM AUTHOR]

Multiview video coding (MVC) aims to compress the multiview video through the elimination of video redundancies, where the quality of the reference frame directly affects the compression efficiency. In this paper, we propose a deep virtual reference frame generation method based on a disparity-aware reference frame generation network (DAG-Net) to transform the disparity relationship between different viewpoints and generate a more reliable reference frame. The proposed DAG-Net consists of a multi-level receptive field module, a disparity-aware alignment module, and a fusion reconstruction module. First, a multi-level receptive field module is designed to enlarge the receptive field, and extract the multi-scale deep features of the temporal and inter-view reference frames. Then, a disparity-aware alignment module is proposed to learn the disparity relationship, and perform disparity shift on the inter-view reference frame to align it with the temporal reference frame. Finally, a fusion reconstruction module is utilized to fuse the complementary information and generate a more reliable virtual reference frame. Experiments demonstrate that the proposed reference frame generation method achieves superior performance for multiview video coding. [ABSTRACT FROM AUTHOR]

We consider the problem of obtaining image quality representations in a self-supervised manner. We use prediction of distortion type and degree as an auxiliary task to learn features from an unlabeled image dataset containing a mixture of synthetic and realistic distortions. We then train a deep Convolutional Neural Network (CNN) using a contrastive pairwise objective to solve the auxiliary problem. We refer to the proposed training framework and resulting deep IQA model as the CONTRastive Image QUality Evaluator (CONTRIQUE). During evaluation, the CNN weights are frozen and a linear regressor maps the learned representations to quality scores in a No-Reference (NR) setting. We show through extensive experiments that CONTRIQUE achieves competitive performance when compared to state-of-the-art NR image quality models, even without any additional fine-tuning of the CNN backbone. The learned representations are highly robust and generalize well across images afflicted by either synthetic or authentic distortions. Our results suggest that powerful quality representations with perceptual relevance can be obtained without requiring large labeled subjective image quality datasets. The implementations used in this paper are available at https://github.com/pavancm/CONTRIQUE. [ABSTRACT FROM AUTHOR]

Due to the increased popularity of augmented and virtual reality experiences, the interest in capturing high-resolution real-world point clouds has never been higher. Loss of details and irregularities in point cloud geometry can occur during the capturing, processing, and compression pipeline. It is essential to address these challenges by being able to upsample a low Level-of-Detail (LoD) point cloud into a high LoD point cloud. Current upsampling methods suffer from several weaknesses in handling point cloud upsampling, especially in dense real-world photo-realistic point clouds. In this paper, we present a novel geometry upsampling technique, PU-Dense, which can process a diverse set of point clouds including synthetic mesh-based point clouds, real-world high-resolution point clouds, real-world indoor LiDAR scanned objects, as well as outdoor dynamically acquired LiDAR-based point clouds. PU-Dense employs a 3D multiscale architecture using sparse convolutional networks that hierarchically reconstruct an upsampled point cloud geometry via progressive rescaling and multiscale feature extraction. The framework employs a UNet type architecture that downscales the point cloud to a bottleneck and then upscales it to a higher level-of-detail (LoD) point cloud. PU-Dense introduces a novel Feature Extraction Unit that incorporates multiscale spatial learning by employing filters at multiple sampling rates and receptive fields. The architecture is memory efficient and is driven by a binary voxel occupancy classification loss that allows it to process high-resolution dense point clouds with millions of points during inference time. Qualitative and quantitative experimental results show that our method significantly outperforms the state-of-the-art approaches by a large margin while having much lower inference time complexity. We further test our dataset on high-resolution photo-realistic datasets. In addition, our method can handle noisy data well. We further show that our approach is memory efficient compared to the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Deep self-expressiveness-based subspace clustering methods have demonstrated effectiveness. However, existing works only consider the attribute information to conduct the self-expressiveness, limiting the clustering performance. In this paper, we propose a novel adaptive attribute and structure subspace clustering network (AASSC-Net) to simultaneously consider the attribute and structure information in an adaptive graph fusion manner. Specifically, we first exploit an auto-encoder to represent input data samples with latent features for the construction of an attribute matrix. We also construct a mixed signed and symmetric structure matrix to capture the local geometric structure underlying data samples. Then, we perform self-expressiveness on the constructed attribute and structure matrices to learn their affinity graphs separately. Finally, we design a novel attention-based fusion module to adaptively leverage these two affinity graphs to construct a more discriminative affinity graph. Extensive experimental results on commonly used benchmark datasets demonstrate that our AASSC-Net significantly outperforms state-of-the-art methods. In addition, we conduct comprehensive ablation studies to discuss the effectiveness of the designed modules. The code is publicly available at https://github.com/ZhihaoPENG-CityU/AASSC-Net. [ABSTRACT FROM AUTHOR]

The difficulties of obtaining sufficient labeled samples have always been one of the factors hindering deep learning models from obtaining high accuracy in hyperspectral image (HSI) classification. To reduce the dependence of deep learning models on training samples, meta learning methods have been introduced, effectively improving the classification accuracy in small sample set scenarios. However, the existing methods based on meta learning still need to construct a labeled source data set with several pre-collected HSIs, and must utilize a large number of labeled samples for meta-training, which is actually time-consuming and labor-intensive. To solve this problem, this paper proposes a novel unsupervised meta learning method with multiview constraints for HSI small sample set classification. Specifically, the proposed method first builds an unlabeled source data set using unlabeled HSIs. Then, multiple spatial-spectral multiview features of each unlabeled sample are generated to construct tasks for unsupervised meta learning. Finally, the designed residual relation network is used for meta-training and small sample set classification based on the voting strategy. Compared with existing supervised meta learning methods for HSI classification, our method can only utilize HSIs without any label for unsupervised meta learning, which significantly reduces the number of requisite labeled samples in the whole classification process. To verify the effectiveness of the proposed method, extensive experiments are carried out on 8 public HSIs in the cross-domain and in-domain classification scenarios. The statistical results demonstrate that, compared with existing supervised meta learning methods and other advanced classification models, the proposed method can achieve competitive or better classification performance in small sample set scenarios. [ABSTRACT FROM AUTHOR]

The practical task of Automatic Check-Out (ACO) is to accurately predict the presence and count of each product in an arbitrary product combination. Beyond the large-scale and the fine-grained nature of product categories as its main challenges, products are always continuously updated in realistic check-out scenarios, which is also required to be solved in an ACO system. Previous work in this research line almost depends on the supervisions of labor-intensive bounding boxes of products by performing a detection paradigm. While, in this paper, we propose a Self-Supervised Multi-Category Counting (S2MC2) network to leverage the point-level supervisions of products in check-out images to both lower the labeling cost and be able to return ACO predictions in a class incremental setting. Specifically, as a backbone, our S2MC2 is built upon a counting module in a class-agnostic counting fashion. Also, it consists of several crucial components including an attention module for capturing fine-grained patterns and a domain adaptation module for reducing the domain gap between single product images as training and check-out images as test. Furthermore, a self-supervised approach is utilized in S2MC2 to initialize the parameters of its backbone for better performance. By conducting comprehensive experiments on the large-scale automatic check-out dataset RPC, we demonstrate that our proposed S2MC2 achieves superior accuracy in both traditional and incremental settings of ACO tasks over the competing baselines. [ABSTRACT FROM AUTHOR]

The success of current deep saliency models heavily depends on large amounts of annotated human fixation data to fit the highly non-linear mapping between the stimuli and visual saliency. Such fully supervised data-driven approaches are annotation-intensive and often fail to consider the underlying mechanisms of visual attention. In contrast, in this paper, we introduce a model based on various cognitive theories of visual saliency, which learns visual attention patterns in a weakly supervised manner. Our approach incorporates insights from cognitive science as differentiable submodules, resulting in a unified, end-to-end trainable framework. Specifically, our model encapsulates the following important components motivated from biological vision. (a) As scene semantics are closely related to visually attentive regions, our model encodes discriminative spatial information for scene understanding through spatial visual semantics embedding. (b) To model the objectness factors in visual attention deployment, we incorporate object-level semantics embedding and object relation information. (c) Considering the “winner-take-all” mechanism in visual stimuli processing, we model the competition mechanism among objects with softmax based neural attention. (d) Lastly, a conditional center prior is learned to mimic the spatial distribution bias of visual attention. Furthermore, we propose novel loss functions to utilize supervision cues from image-level semantics, saliency prior knowledge, and self-information compression. Experiments show that our method achieves promising results, and even outperforms many of its fully supervised counterparts. Overall, our weakly supervised saliency method makes an essential step towards reducing the annotation budget of current approaches, as well as providing a more comprehensive understanding of the visual attention mechanism. Our code is available at: https://github.com/ashleylqx/WeakFixation.git. [ABSTRACT FROM AUTHOR]

Object detection is usually solved by learning a deep architecture involving classification and localization tasks, where feature learning for these two tasks is shared using the same backbone model. Recent works have shown that suitable disentanglement of classification and localization tasks has the great potential to improve performance of object detection. Despite the promising performance, existing feature disentanglement methods usually suffer from two limitations. First, most of them only focus on the disentangled proposals or predication heads for classification and localization tasks after RPN. While little consideration has been given to that the features for these two different tasks actually are obtained by a shared backbone model before RPN. Second, they are suggested for two-stage objectors and are not applicable to one-stage methods. To overcome these limitations, this paper presents a novel fully task-specific feature learning method for one-stage object detection. Specifically, our method first learns disentangled features for classification and localization tasks using two separated backbone models, where auxiliary classification and localization heads are inserted at the end of the two backbone models for providing a fully task-specific features for classification and localization. Then, a feature interaction module is developed for aligning and fusing task-specific features, which are further used to produce the final detection result. Experiments on MS COCO show that our proposed method (dubbed CrabNet) can achieve clear improvement over counterparts with increasing limited inference time, while performing favorably against state-of-the-arts. [ABSTRACT FROM AUTHOR]

Unsupervised active learning has become an active research topic in the machine learning and computer vision communities, whose goal is to choose a subset of representative samples to be labeled in an unsupervised setting. Most of existing approaches rely on shallow linear models by assuming that each sample can be well approximated by the span (i.e., the set of all linear combinations) of the selected samples, and then take these selected samples as the representative ones for manual labeling. However, the data do not necessarily conform to the linear models in many real-world scenarios, and how to model nonlinearity of data often becomes the key point of unsupervised active learning. Moreover, the existing works often aim to well reconstruct the whole dataset, while ignore the important cluster structure, especially for imbalanced data. In this paper, we present a novel deep unsupervised active learning framework. The proposed method can explicitly learn a nonlinear embedding to map each input into a latent space via a deep neural network, and introduce a selection block to select the representative samples in the learnt latent space through a self-supervised learning strategy. In the selection block, we aim to not only preserve the global structure of the data, but also capture the cluster structure of the data in order to well handle the data imbalance issue during sample selection. Meanwhile, we take advantage of the clustering result to provide self-supervised information to guide the above processes. Finally, we attempt to preserve the local structure of the data, such that the data embedding becomes more precise and the model performance can be further improved. Extensive experimental results on several publicly available datasets clearly demonstrate the effectiveness of our method, compared with the state-of-the-arts. [ABSTRACT FROM AUTHOR]

Deep object detection models trained on clean images may not generalize well on degraded images due to the well-known domain shift issue. This hinders their application in real-life scenarios such as video surveillance and autonomous driving. Though domain adaptation methods can adapt the detection model from a labeled source domain to an unlabeled target domain, they struggle in dealing with open and compound degradation types. In this paper, we attempt to address this problem in the context of object detection by proposing a robust object Detector via Adversarial Novel Style Exploration (DANSE). Technically, DANSE first disentangles images into domain-irrelevant content representation and domain-specific style representation under an adversarial learning framework. Then, it explores the style space to discover diverse novel degradation styles that are complementary to those of the target domain images by leveraging a novelty regularizer and a diversity regularizer. The clean source domain images are transferred into these discovered styles by using a content-preserving regularizer to ensure realism. These transferred source domain images are combined with the target domain images and used to train a robust degradation-agnostic object detection model via adversarial domain adaptation. Experiments on both synthetic and real benchmark scenarios confirm the superiority of DANSE over state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Video Summarization (VS) has become one of the most effective solutions for quickly understanding a large volume of video data. Dictionary selection with self representation and sparse regularization has demonstrated its promise for VS by formulating the VS problem as a sparse selection task on video frames. However, existing dictionary selection models are generally designed only for data reconstruction, which results in the neglect of the inherent structured information among video frames. In addition, the sparsity commonly constrained by $L_{2,1}$ norm is not strong enough, which causes the redundancy of keyframes, i.e., similar keyframes are selected. Therefore, to address these two issues, in this paper we propose a general framework called graph convolutional dictionary selection with $L_{2,p}$ ($0< p\leq 1$) norm (GCDS $_{2,p}$) for both keyframe selection and skimming based summarization. Firstly, we incorporate graph embedding into dictionary selection to generate the graph embedding dictionary, which can take the structured information depicted in videos into account. Secondly, we propose to use $L_{2,p}$ ($0< p\leq 1$) norm constrained row sparsity, in which $p$ can be flexibly set for two forms of video summarization. For keyframe selection, $0< p< 1$ can be utilized to select diverse and representative keyframes; and for skimming, $p=1$ can be utilized to select key shots. In addition, an efficient iterative algorithm is devised to optimize the proposed model, and the convergence is theoretically proved. Experimental results including both keyframe selection and skimming based summarization on four benchmark datasets demonstrate the effectiveness and superiority of the proposed method. [ABSTRACT FROM AUTHOR]

Hyperspectral imagery with very high spectral resolution provides a new insight for subtle nuances identification of similar substances. However, hyperspectral target detection faces significant challenges of intraclass dissimilarity and interclass similarity due to the unavoidable interference caused by atmosphere, illumination, and sensor noise. In order to effectively alleviate these spectral inconsistencies, this paper proposes a novel target detection method without strict assumptions on data distribution based on an unconstrained linear mixture model and deep learning. Our proposed detector firstly reduces interference via a specifically designed deep-learning-based hierarchical denoising autoencoder, and then carries out accurate detection with a two-step subspace projection, aiming at background suppression and target enhancement. Additionally, to generate representative background and reliable target samples required in the detection procedure, an efficient spatial-spectral unified endmember extraction method has been developed. Performance comparison with several state-of-the-art detection methods and further analysis on four real-world hyperspectral images demonstrate the effectiveness and efficiency of our proposed target detector. [ABSTRACT FROM AUTHOR]

In this paper, we aim to explore the fine-grained perception ability of deep models for the newly proposed scene sketch semantic segmentation task. Scene sketches are abstract drawings containing multiple related objects. It plays a vital role in daily communication and human-computer interaction. The study has only recently started due to a main obstacle of the absence of large-scale datasets. The currently available dataset SketchyScene is composed of clip art-style edge maps, which lacks abstractness and diversity. To drive further research, we contribute two new large-scale datasets based on real hand-drawn object sketches. A general automatic scene sketch synthesis process is developed to assist with new dataset composition. Furthermore, we propose to enhancing local detail perception in deep models to realize accurate stroke-oriented scene sketch segmentation. Due to the inherent differences between hand-drawn sketches and natural images, extreme low-level local features of strokes are incorporated to improve detail discrimination. Stroke masks are also integrated into model training to guide the learning attention. Extensive experiments are conducted on three large-scale scene sketch datasets. Our method achieves state-of-the-art performance under four evaluation metrics and yields meaningful interpretability via visual analytics. [ABSTRACT FROM AUTHOR]

Convolutional dictionary learning has become increasingly popular in signal and image processing for its ability to overcome the limitations of traditional patch-based dictionary learning. Although most studies on convolutional dictionary learning mainly focus on the unimodal case, real-world image processing tasks usually involve images from multiple modalities, e.g., visible and near-infrared (NIR) images. Thus, it is necessary to explore convolutional dictionary learning across different modalities. In this paper, we propose a novel multi-modal convolutional dictionary learning algorithm, which efficiently correlates different image modalities and fully considers neighborhood information at the image level. In this model, each modality is represented by two convolutional dictionaries, in which one dictionary is for common feature representation and the other is for unique feature representation. The model is constrained by the requirement that the convolutional sparse representations (CSRs) for the common features should be the same across different modalities, considering that these images are captured from the same scene. We propose a new training method based on the alternating direction method of multipliers (ADMM) to alternatively learn the common and unique dictionaries in the discrete Fourier transform (DFT) domain. We show that our model converges in less than 20 iterations between the convolutional dictionary updating and the CSRs calculation. The effectiveness of the proposed dictionary learning algorithm is demonstrated on various multimodal image processing tasks, achieves better performance than both dictionary learning methods and deep learning based methods with limited training data. [ABSTRACT FROM AUTHOR]

In this paper, we make the first benchmark effort to elaborate on the superiority of using RAW images in the low light enhancement and develop a novel alternative route to utilize RAW images in a more flexible and practical way. Inspired by a full consideration on the typical image processing pipeline, we are inspired to develop a new evaluation framework, Factorized Enhancement Model (FEM), which decomposes the properties of RAW images into measurable factors and provides a tool for exploring how properties of RAW images affect the enhancement performance empirically. The empirical benchmark results show that the Linearity of data and Exposure Time recorded in meta-data play the most critical role, which brings distinct performance gains in various measures over the approaches taking the sRGB images as input. With the insights obtained from the benchmark results in mind, a RAW-guiding Exposure Enhancement Network (REENet) is developed, which makes trade-offs between the advantages and inaccessibility of RAW images in real applications in a way of using RAW images only in the training phase. REENet projects sRGB images into linear RAW domains to apply constraints with corresponding RAW images to reduce the difficulty of modeling training. After that, in the testing phase, our REENet does not rely on RAW images. Experimental results demonstrate not only the superiority of REENet to state-of-the-art sRGB-based methods and but also the effectiveness of the RAW guidance and all components. [ABSTRACT FROM AUTHOR]

Different from the object motion blur, the defocus blur is caused by the limitation of the cameras’ depth of field. The defocus amount can be characterized by the parameter of point spread function and thus forms a defocus map. In this paper, we propose a new network architecture called Defocus Image Deblurring Auxiliary Learning Net (DID-ANet), which is specifically designed for single image defocus deblurring by using defocus map estimation as auxiliary task to improve the deblurring result. To facilitate the training of the network, we build a novel and large-scale dataset for single image defocus deblurring, which contains the defocus images, the defocus maps and the all-sharp images. To the best of our knowledge, the new dataset is the first large-scale defocus deblurring dataset for training deep networks. Moreover, the experimental results demonstrate that the proposed DID-ANet outperforms the state-of-the-art methods for both tasks of defocus image deblurring and defocus map estimation, both quantitatively and qualitatively. The dataset, code, and model is available on GitHub: https://github.com/xytmhy/DID-ANet-Defocus-Deblurring. [ABSTRACT FROM AUTHOR]

Adversarial examples are inputs with imperceptible perturbations that easily mislead deep neural networks (DNNs). Recently, adversarial patch, with noise confined to a small and localized patch, has emerged for its easy feasibility in real-world scenarios. However, existing strategies failed to generate adversarial patches with strong generalization ability due to the ignorance of the inherent biases of models. In other words, the adversarial patches are always input-specific and fail to attack images from all classes or different models, especially unseen classes and black-box models. To address the problem, this paper proposes a bias-based framework to generate universal adversarial patches with strong generalization ability, which exploits the perceptual bias and attentional bias to improve the attacking ability. Regarding the perceptual bias, since DNNs are strongly biased towards textures, we exploit the hard examples which convey strong model uncertainties and extract a textural patch prior from them by adopting the style similarities. The patch prior is closer to decision boundaries and would promote attacks across classes. As for the attentional bias, motivated by the fact that different models share similar attention patterns towards the same image, we exploit this bias by confusing the model-shared similar attention patterns. Thus, the generated adversarial patches can obtain stronger transferability among different models. Taking Automatic Check-out (ACO) as the typical scenario, extensive experiments including white-box/black-box settings in both digital-world (RPC, the largest ACO related dataset) and physical-world scenario (Taobao and JD, the world’s largest online shopping platforms) are conducted. Experimental results demonstrate that our proposed framework outperforms state-of-the-art adversarial patch attack methods. [ABSTRACT FROM AUTHOR]