Search

Highlights • A self-supervised learning which does not require human annotations for training CNN. • Learning the correct arrangement of object proposals to represent an image by CNN. • Demonstrating the advantage of our model by applying it to PASCAL VOC datasets. • Application to other vision tasks including image retrieval and semantic matching. Abstract Learning visual representations plays an important role in computer vision and machine learning applications. It facilitates a model to understand and perform high-level tasks intelligently. A common approach for learning visual representations is supervised one which requires a huge amount of human annotations to train the model. This paper presents a self-supervised approach which learns visual representations from input images without human annotations. We learn the correct arrangement of object proposals to represent an image using a convolutional neural network (CNN) without any manual annotations. We hypothesize that the network trained for solving this problem requires the embedding of semantic visual representations. Unlike existing approaches that use uniformly sampled patches, we relate object proposals that contain prominent objects and object parts. More specifically, we discover the representation that considers overlap, inclusion, and exclusion relationship of proposals as well as their relative position. This allows focusing on potential objects and parts rather than on clutter. We demonstrate that our model outperforms existing self-supervised learning methods and can be used as a generic feature extractor by applying it to object detection, classification, action recognition, image retrieval, and semantic matching tasks. [ABSTRACT FROM AUTHOR]

Most variational formulations for structure-texture image decomposition force the structure images to have small norm in some functional spaces and to share a common notion of edges, i.e., large-gradients or large-intensity differences. However, such a definition makes it difficult to distinguish structure edges from oscillations that have fine spatial scale but high contrast. In this paper, we introduce a new model by learning deep variational priors for structure images without explicit training data. An alternating direction method of a multiplier algorithm and its modular structure are adopted to plug deep variational priors into an iterative smoothing process. The central observations are that convolution neural networks (CNNs) can replace the total variation prior, and are indeed powerful to capture the natures of structure and texture. We show that our learned priors using CNNs successfully differentiate high-amplitude details from structure edges, and avoid halo artifacts. Different from previous data-driven smoothing schemes, our formulation provides another degree of freedom to produce continuous smoothing effects. Experimental results demonstrate the effectiveness of our approach on various computational photography and image processing applications, including texture removal, detail manipulation, HDR tone-mapping, and non-photorealistic abstraction. [ABSTRACT FROM AUTHOR]

Finding image correspondences remains a challenging problem in the presence of intra-class variations and large changes in scene layout. Semantic flow methods are designed to handle images depicting different instances of the same object or scene category. We introduce a novel approach to semantic flow, dubbed proposal flow, that establishes reliable correspondences using object proposals. Unlike prevailing semantic flow approaches that operate on pixels or regularly sampled local regions, proposal flow benefits from the characteristics of modern object proposals, that exhibit high repeatability at multiple scales, and can take advantage of both local and geometric consistency constraints among proposals. We also show that the corresponding sparse proposal flow can effectively be transformed into a conventional dense flow field. We introduce two new challenging datasets that can be used to evaluate both general semantic flow techniques and region-based approaches such as proposal flow. We use these benchmarks to compare different matching algorithms, object proposals, and region features within proposal flow, to the state of the art in semantic flow. This comparison, along with experiments on standard datasets, demonstrates that proposal flow significantly outperforms existing semantic flow methods in various settings. [ABSTRACT FROM AUTHOR]

Filtering images using a guidance signal, a process called guided or joint image filtering, has been used in various tasks in computer vision and computational photography, particularly for noise reduction and joint upsampling. This uses an additional guidance signal as a structure prior, and transfers the structure of the guidance signal to an input image, restoring noisy or altered image structure. The main drawbacks of such a data-dependent framework are that it does not consider structural differences between guidance and input images, and that it is not robust to outliers. We propose a novel SD (for static/dynamic) filter to address these problems in a unified framework, and jointly leverage structural information from guidance and input images. Guided image filtering is formulated as a nonconvex optimization problem, which is solved by the majorize-minimization algorithm. The proposed algorithm converges quickly while guaranteeing a local minimum. The SD filter effectively controls the underlying image structure at different scales, and can handle a variety of types of data from different sensors. It is robust to outliers and other artifacts such as gradient reversal and global intensity shift, and has good edge-preserving smoothing properties. We demonstrate the flexibility and effectiveness of the proposed SD filter in a variety of applications, including depth upsampling, scale-space filtering, texture removal, flash/non-flash denoising, and RGB/NIR denoising. [ABSTRACT FROM AUTHOR]

Interactive image segmentation has remained an active research topic in image processing and graphics, since the user intention can be incorporated to enhance the performance. It can be employed to mobile devices which now allow user interaction as an input, enabling various applications. Most interactive segmentation methods assume that the initial labels are correctly and carefully assigned to some parts of regions to segment. Inaccurate labels, such as foreground labels in background regions for example, lead to incorrect segments, even by a small number of inaccurate labels, which is not appropriate for practical usage such as mobile application. In this paper, we present an interactive segmentation method that is robust to inaccurate initial labels (scribbles). To address this problem, we propose a structure-aware labeling method using occurrence and co-occurrence probability (OCP) of color values for each initial label in a unified framework. Occurrence probability captures a global distribution of all color values within each label, while co-occurrence one encodes a local distribution of color values around the label. We show that nonlocal regularization together with the OCP enables robust image segmentation to inaccurately assigned labels and alleviates a small-cut problem. We analyze theoretic relations of our approach to other segmentation methods. Intensive experiments with synthetic and manual labels show that our approach outperforms the state of the art. [ABSTRACT FROM AUTHOR]

A nonlinear disparity remapping scheme is presented to enhance the visual comfort of stereoscopic videos. The stereoscopic video is analyzed for predicting a degree of fatigue with the viewpoint of three factors: 1) spatial frequency; 2) disparity magnitude; and 3) disparity motion. The degree of fatigue is then estimated in a local manner. It can be visualized as an index map so-called a “visual fatigue map,” and an overall fatigue score is obtained by pooling the visual fatigue map. Based on this information, a nonlinear remapping operator is generated in two phases: 1) disparity range adaptation and 2) operator nonlinearization. First, a disparity range is automatically adjusted according to the determined overall fatigue score. Second, rather than linearly adjusting the disparity range of an original video to the determined disparity range, a nonlinear remapping operator is constructed in a manner that the disparity range of inducible problematic region is compressed, while that of comfortable region is stretched. The proposed scheme is verified via subjective evaluations where visual fatigue and depth sensation are compared among original videos, linearly remapped videos, and nonlinearly remapped videos. Experimental results show that the nonlinearly remapped videos provide more comfort than the linearly remapped videos without losing depth sensation. [ABSTRACT FROM AUTHOR]

This paper presents a depth superresolution (SR) method that uses both of a low-resolution (LR) depth image and a high-resolution (HR) intensity image. We formulate depth SR as a graph-based transduction problem. In particular, the HR intensity image is represented as an undirected graph, in which pixels are characterized as vertices, and their relations are encoded as an affinity function. When the vertices initially labeled with certain depth hypotheses (from the LR depth image) are regarded as input queries, all the vertices are scored with respect to the relevances to these queries by a classifying function. Each vertex is then labeled with the depth hypothesis that receives the highest relevance score. We design the classifying function by considering the local and global structures of the HR intensity image. This approach enables us to address a depth bleeding problem that typically appears in current depth SR methods. Furthermore, input queries are assigned in a probabilistic manner, making depth SR robust to noisy depth measurements. We also analyze existing depth SR methods in the context of transduction, and discuss their theoretic relations. Intensive experiments demonstrate the superiority of the proposed method over state-of-the-art methods both qualitatively and quantitatively. [ABSTRACT FROM AUTHOR]

A robust similarity measure called the Mahalanobis distance cross-correlation (MDCC) is proposed for illumination-invariant stereo matching, which uses a local color distribution within support windows. It is shown that the Mahalanobis distance between the color itself and the average color is preserved under affine transformation. The MDCC converts pixels within each support window into the Mahalanobis distance transform (MDT) space. The similarity between MDT pairs is then computed using the cross-correlation with an asymmetric weight function based on the Mahalanobis distance. The MDCC considers correlation on cross-color channels, thus providing robustness to affine illumination variation. Experimental results show that the MDCC outperforms state-of-the-art similarity measures in terms of stereo matching for image pairs taken under different illumination conditions. [ABSTRACT FROM PUBLISHER]

In this paper, the origin of random walk with restart (RWR) and its generalization are described. It is well known that the random walk (RW) and the anisotropic diffusion models share the same energy functional, i.e., the former provides a steady-state solution and the latter gives a flow solution. In contrast, the theoretical background of the RWR scheme is different from that of the diffusion-reaction equation, although the restarting term of the RWR plays a role similar to the reaction term of the diffusion-reaction equation. The behaviors of the two approaches with respect to outliers reveal that they possess different attributes in terms of data propagation. This observation leads to the derivation of a new energy functional, where both volumetric heat capacity and thermal conductivity are considered together, and provides a common framework that unifies both the RW and the RWR approaches, in addition to other regularization methods. The proposed framework allows the RWR to be generalized (GRWR) in semilocal and nonlocal forms. The experimental results demonstrate the superiority of GRWR over existing regularization approaches in terms of depth map up-sampling and interactive image segmentation. [ABSTRACT FROM PUBLISHER]

Joint image filters are used to transfer structural details from a guidance picture used as a prior to a target image, in tasks such as enhancing spatial resolution and suppressing noise. Previous methods based on convolutional neural networks (CNNs) combine nonlinear activations of spatially-invariant kernels to estimate structural details and regress the filtering result. In this paper, we instead learn explicitly sparse and spatially-variant kernels. We propose a CNN architecture and its efficient implementation, called the deformable kernel network (DKN), that outputs sets of neighbors and the corresponding weights adaptively for each pixel. The filtering result is then computed as a weighted average. We also propose a fast version of DKN that runs about seventeen times faster for an image of size 640 × 480 . We demonstrate the effectiveness and flexibility of our models on the tasks of depth map upsampling, saliency map upsampling, cross-modality image restoration, texture removal, and semantic segmentation. In particular, we show that the weighted averaging process with sparsely sampled 3 × 3 kernels outperforms the state of the art by a significant margin in all cases. [ABSTRACT FROM AUTHOR]

We present a descriptor, called fully convolutional self-similarity (FCSS), for dense semantic correspondence. Unlike traditional dense correspondence approaches for estimating depth or optical flow, semantic correspondence estimation poses additional challenges due to intra-class appearance and shape variations among different instances within the same object or scene category. To robustly match points across semantically similar images, we formulate FCSS using local self-similarity (LSS), which is inherently insensitive to intra-class appearance variations. LSS is incorporated through a proposed convolutional self-similarity (CSS) layer, where the sampling patterns and the self-similarity measure are jointly learned in an end-to-end and multi-scale manner. Furthermore, to address shape variations among different object instances, we propose a convolutional affine transformer (CAT) layer that estimates explicit affine transformation fields at each pixel to transform the sampling patterns and corresponding receptive fields. As training data for semantic correspondence is rather limited, we propose to leverage object candidate priors provided in most existing datasets and also correspondence consistency between object pairs to enable weakly-supervised learning. Experiments demonstrate that FCSS significantly outperforms conventional handcrafted descriptors and CNN-based descriptors on various benchmarks. [ABSTRACT FROM AUTHOR]

Establishing dense correspondences between multiple images is a fundamental task in many applications. However, finding a reliable correspondence between multi-modal or multi-spectral images still remains unsolved due to their challenging photometric and geometric variations. In this paper, we propose a novel dense descriptor, called dense adaptive self-correlation (DASC), to estimate dense multi-modal and multi-spectral correspondences. Based on an observation that self-similarity existing within images is robust to imaging modality variations, we define the descriptor with a series of an adaptive self-correlation similarity measure between patches sampled by a randomized receptive field pooling, in which a sampling pattern is obtained using a discriminative learning. The computational redundancy of dense descriptors is dramatically reduced by applying fast edge-aware filtering. Furthermore, in order to address geometric variations including scale and rotation, we propose a geometry-invariant DASC (GI-DASC) descriptor that effectively leverages the DASC through a superpixel-based representation. For a quantitative evaluation of the GI-DASC, we build a novel multi-modal benchmark as varying photometric and geometric conditions. Experimental results demonstrate the outstanding performance of the DASC and GI-DASC in many cases of dense multi-modal and multi-spectral correspondences. [ABSTRACT FROM AUTHOR]

A vision based real-time rear obstacle detection system is one of the most essential technologies, which can be used in many applications such as a parking assistance systems and intelligent vehicles. Although disparity is a useful feature for detecting obstacles, estimating a correct disparity map is a hard problem due to the matching ambiguity and noise sensitivity, especially in homogeneous regions. To overcome these problems, we leverage reliable disparities only for obstacle detection. A reliability factor is introduced to measure an inhomogeneity of the regions quantitatively. It is computed at each superpixel to consider the noise sensitivity of pixel-wise gradients and to assign similar reliability value within a same object. It includes two major components: firstly, In a feature extraction and combining stage, we extract three features from stereo images such as disparity, superpixel segments and pixel-wise gradient and compute the reliability of disparity from superpixel segments and the pixel-wise gradient. Secondly, In an obstacle detection stage, a disparity feature with reliability votes for localizing obstacles and dominant candidates in voting map are selected as initial obstacle region. The initial obstacle regions are expanded into their neighbor superpixels based on CIELAB color similarity and distance similarity between superpixels. Experimental results show satisfactory performance under various real parking environments. Its detection rate is at least 4% higher than those of other existing methods, and its false detection rate is more than 10% lower and thus, can be used for parking assistance system. [ABSTRACT FROM AUTHOR]

We address the problem of person re-identification (reID), that is, retrieving person images from a large dataset, given a query image of the person of interest. A key challenge is to learn person representations robust to intra-class variations, as different persons could have the same attribute, and persons’ appearances look different, e.g., with viewpoint changes. Recent reID methods focus on learning person features discriminative only for a particular factor of variations (e.g., human pose), which also requires corresponding supervisory signals (e.g., pose annotations). To tackle this problem, we propose to factorize person images into identity-related and -unrelated features. Identity-related features contain information useful for specifying a particular person (e.g., clothing), while identity-unrelated ones hold other factors (e.g., human pose). To this end, we propose a new generative adversarial network, dubbed identity shuffle GAN (IS-GAN). It disentangles identity-related and -unrelated features from person images through an identity-shuffling technique that exploits identification labels alone without any auxiliary supervisory signals. We restrict the distribution of identity-unrelated features, or encourage the identity-related and -unrelated features to be uncorrelated, facilitating the disentanglement process. Experimental results validate the effectiveness of IS-GAN, showing state-of-the-art performance on standard reID benchmarks, including Market-1501, CUHK03 and DukeMTMC-reID. We further demonstrate the advantages of disentangling person representations on a long-term reID task, setting a new state of the art on a Celeb-reID dataset. Our code and models are available online: https://cvlab-yonsei.github.io/projects/ISGAN/. [ABSTRACT FROM AUTHOR]

We address the problem of semantic correspondence, that is, establishing a dense flow field between images depicting different instances of the same object or scene category. We propose to use images annotated with binary foreground masks and subjected to synthetic geometric deformations to train a convolutional neural network (CNN) for this task. Using these masks as part of the supervisory signal provides an object-level prior for the semantic correspondence task and offers a good compromise between semantic flow methods, where the amount of training data is limited by the cost of manually selecting point correspondences, and semantic alignment ones, where the regression of a single global geometric transformation between images may be sensitive to image-specific details such as background clutter. We propose a new CNN architecture, dubbed SFNet, which implements this idea. It leverages a new and differentiable version of the argmax function for end-to-end training, with a loss that combines mask and flow consistency with smoothness terms. Experimental results demonstrate the effectiveness of our approach, which significantly outperforms the state of the art on standard benchmarks. [ABSTRACT FROM AUTHOR]

Person search involves localizing and re-identifying persons of interest captured by multiple, non-overlapping cameras. Recent approaches to person search are typically built on object detection frameworks to learn joint person representations for detection and re-identification. To this end, the features extracted from pedestrian proposals are projected onto a unit hypersphere using L2 normalization, and positive proposals that sufficiently overlap with the ground truth are equally incorporated for training by exploiting an external lookup table (LUT). We have found that (1) using the L2 normalization technique, without considering feature distributions, can degenerate the discriminative power of person representations, (2) positive proposals often depict distracting details, such as background clutter and person overlaps, and (3) person features in the LUT are not often updated during training. To address these limitations, we propose a novel framework for person search, dubbed PLoPS, using a prototypical normalization layer, ProtoNorm, that calibrates features while considering the long-tail distribution across person IDs. PLoPS also entails a localization-aware learning scheme that prioritizes better-aligned proposals w.r.t the ground truth. We further introduce a LUT calibration technique to continuously adjust the person features in the LUT. Experimental results and analysis on standard benchmarks demonstrate the effectiveness of PLoPS. • A novel normalization layer, ProtoNorm, considering the class imbalance problem across person IDs. • A localization-aware objective function for learning discriminative features for person search. • A LUT calibration technique that adjusts person features in the LUT during training. • We set a new state of the art on standard benchmarks for person search. [ABSTRACT FROM AUTHOR]