CVPR 2015 Papers

Graph-theoretical methods have successfully provided semantic and structural interpretations of images and videos. A recent paper introduced a pattern-theoretic approach that allows construction of flexible graphs for representing interactions of actors with objects and inference is accomplished by an efficient annealing algorithm. Actions and objects are termed generators and their interactions are termed bonds; together they form high-probability configurations, or interpretations, of observed scenes. This work and other structural methods have generally been limited to analyzing short videos involving isolated actions. Here we provide an extension that uses additional temporal bonds across individual actions to enable semantic interpretations of longer videos. Longer temporal connections improve scene interpretations as they help discard (temporally) local solutions in favor of globally superior ones. Using this extension, we demonstrate improvements in understanding longer videos, compared to individual interpretations of non-overlapping time segments. We verified the success of our approach by generating interpretations for more than 700 video segments from the YouCook data set, with intricate videos that exhibit cluttered background, scenarios of occlusion, viewpoint variations and changing conditions of illumination. Interpretations for long video segments were able to yield performance increases of about 70 and, in addition, proved to be more robust to different severe scenarios of classification errors.

In this paper we show that multiple object tracking (MOT) can be formulated in a framework, where the detection and data-association are performed simultaneously. Our method allows us to overcome the confinements of data association based MOT approaches; where the performance is dependent on the object detection results provided at input level. At the core of our method lies structured learning which learns a model for each target and infers the best location of all targets simultaneously in a video clip. The inference of our structured learning is done through a new Target Identity-aware Network Flow (TINF), where each node in the network encodes the probability of each target identity belonging to that node. The proposed Lagrangian relaxation optimization finds the high quality solution to the network. During optimization a soft spatial constraint is enforced between the nodes of the graph which helps reducing the ambiguity caused by nearby targets with similar appearance in crowded scenarios. We show that automatically detecting and tracking targets in a single framework can help resolve the ambiguities due to frequent occlusion and heavy articulation of targets. Our experiments involve challenging yet distinct datasets and show that our method can achieve results better than the state-of-art.

The only goal of the abstract is to the answer the question: why should I read this paper? In this paper we study the use of convolutional neural networks (convnets) for the task of pedestrian detection. Despite their recent diverse successes, convnets historically underperform compared to other pedestrian detectors. We deliberately omit explicitly modelling the problem into the network (e.g. parts or occlusion modelling) and show that we can reach competitive performance without bells and whistles. In a wide range of experiments we analyse small and big convnets, their architectural choices, parameters, and the influence of different training data, including pre-training on surrogate tasks. We present the best convnet detectors on the Caltech and KITTI dataset. On Caltech our convnets reach top performance both for the Caltech1x and Caltech10x training setup. Using additional data at training time our strongest convnet model is competitive to detectors that use instead additional data at test time.

Video summarization is a challenging problem in part because knowing which part of a video is important requires prior knowledge about its main topic. We present TVSum, an unsupervised video summarization framework that uses title-based image search results to find visually important shots. We observe that a video title is often carefully chosen to be maximally descriptive of its main topic, and hence images related to the title can serve as a proxy for important visual concepts of the main topic. However, because titles are free-formed, unconstrained, and often written ambiguously, images searched using the title can contain noise (images irrelevant to video content) and variance (images of different topics). To deal with this challenge, we developed a novel co-archetypal analysis technique that learns canonical visual concepts shared between video and images, but not in either alone, by finding a joint-factorial representation of two data sets. We introduce a new benchmark dataset, TVSum50, that contains 50 videos and their shot-level importance scores annotated via crowdsourcing. Experimental results on two datasets, SumMe and TVSum50, suggest our approach produces superior quality summaries compared to several recently proposed approaches.

We introduce a learning-based approach to detect repeatable keypoints under drastic imaging changes of weather and lighting conditions to which state-of-the-art keypoint detectors are surprisingly sensitive. We first identify good keypoint candidates in multiple training images taken from the same viewpoint. We then train a regressor to predict a score map whose maxima are those points so that they can be found by simple non-maximum suppression. As there are no standard datasets to test the influence of these kinds of changes, we created our own, which we will make publicly available. We will show that our method significantly outperforms the state-of-the-art methods in such challenging conditions, while still achieving state-of-the-art performance on the untrained standard Oxford dataset.

Recently, a variety of real-world applications have triggered huge demand for techniques that can extract textual information from natural scenes. Therefore, scene text detection and recognition have become active research topics in computer vision. In this work, we investigate the problem of scene text detection from an alternative perspective and propose a novel algorithm for it. Different from traditional methods, which mainly make use of the properties of single characters or strokes, the proposed algorithm exploits the symmetry property of character groups and allows for direct extraction of text lines from natural images. The experiments on the latest ICDAR benchmarks demonstrate that the proposed algorithm achieves state-of-the-art performance. Moreover, compared to conventional approaches, the proposed algorithm shows stronger adaptability to texts in challenging scenarios.

This paper addresses the problem of learning word image representations: given the cropped image of a word, we are interested in finding a descriptive, robust, and compact fixed-length representation. Machine learning techniques can then be supplied with these representations to produce models useful for word retrieval or recognition tasks. Although many works have focused on the machine learning aspect once a global representation has been produced, little work has been devoted to the construction of those base image representations: most works use standard coding and aggregation techniques directly on top of standard computer vision features such as SIFT or HOG. We propose to learn local mid-level features suitable for building word image representations. These features are learnt by leveraging character bounding box annotations on a small set of training images. However, contrary to other approaches that use character bounding box information, our approach does not rely on detecting the individual characters explicitly at testing time. Our local mid-level features can then be aggregated to produce a global word image signature. When pairing these features with the recent word attributes framework of Almaz\'an et al., we obtain results comparable with or better than the state-of-the-art on matching and recognition tasks using global descriptors of only 96 dimensions.

Recently, learning based hashing techniques have attracted broad research interests due to the resulting efficient storage and retrieval of images, videos, documents, etc. However, a major difficulty of learning to hash lies in handling the discrete constraints imposed on the needed hash codes, which typically makes hash optimizations very challenging (NP-hard in general). In this work, we propose a new supervised hashing framework, where the learning objective for hashing is to make the optimal binary hash codes for classification. By introducing an auxiliary variable, we reformulate the objective such that it can be solved substantially efficiently by using a regularization algorithm. One of the key steps in the algorithm is to solve the regularization sub-problem associated with the NP-hard binary optimization. We show that with cyclic coordinate descent, the sub-problem admits an analytical solution. As such, a high-quality discrete solution can eventually be obtained in an efficient computing manner, which enables to tackle massive datasets. We evaluate the proposed approach, dubbed Supervised Discrete Hashing (SDH), on four large image datasets, and demonstrate that SDH outperforms the state-of-the-art hashing methods in large-scale image retrieval.

Descriptor learning has recently drawn increasing attention in computer vision, Existing algorithms are mainly developed for classification rather than for regression which however has recently emerged as a powerful tool to solve a broad range of problems, e.g., head pose estimation. In this paper, we propose a novel supervised descriptor learning (SDL) algorithm to establish a discriminative and compact feature representation for multi-output regression. By formulating as generalized low-rank approximations of matrices with a supervised manifold regularization (SMR), the SDL removes irrelevant and redundant information from raw features by transforming into a low-dimensional space under the supervision of multivariate targets. The obtained discriminative while compact descriptor largely reduces the variability and ambiguity in multi-output regression, and therefore enables more accurate and efficient multivariate estimation. We demonstrate the effectiveness of the proposed SDL algorithm on a representative multi-output regression task: head pose estimation using the benchmark Pointing'04 dataset. Experimental results show that the SDL can achieve high pose estimation accuracy and significantly outperforms state-of-the-art algorithms by an error reduction up to 27.5%. The proposed SDL algorithm provides a general descriptor learning framework in a supervised way for multi-output regression which can largely boost the performance of existing multi-output regression tasks.

In this paper we present an approach for segmenting objects in videos taken in complex scenes with multiple and different targets. The method does not make any specific assumptions about the videos and relies on how objects are perceived by humans according to Gestalt laws. Initially, we rapidly generate a coarse foreground segmentation, which provides predictions about motion regions by analyzing how superpixel segmentation changes in consecutive frames. We then exploit these location priors to refine the initial segmentation by optimizing an energy function based on appearance and perceptual organization, only on regions where motion is observed. We evaluated our method on complex and challenging video sequences and it showed significant performance improvements over recent state-of-the-art methods, being also fast enough to be used for "on-the-fly" processing.

We present in this paper a superpixel segmentation algorithm called Linear Spectral Clustering (LSC), which produces compact and uniform superpixels with low computational costs. Basically, a normalized cuts formulation of the superpixel segmentation is adopted based on a similarity metric that measures the color similarity and space proximity between image pixels. However, instead of using the traditional eigen-based algorithm, we approximate the similarity metric using a kernel function leading to an explicitly mapping of pixel values and coordinates into a high dimensional feature space. We revisit the conclusion that by appropriately weighting each point in this feature space, the objective functions of weighted K-means and normalized cuts share the same optimum point. As such, it is possible to optimize the cost function of normalized cuts by iteratively applying simple K-means clustering in the proposed feature space. LSC is of linear computational complexity and high memory efficiency and is able to preserve global properties of images. Experimental results show that LSC performs equally well or better than state of the art superpixel segmentation algorithms in terms of several commonly used evaluation metrics in image segmentation.

Multi-View-Stereo (MVS) methods aim for the highest detail possible, however, such detail is often not required. In this work, we propose a novel surface reconstruction method based on image edges, superpixels and second-order smoothness constraints, producing meshes comparable to classic MVS surfaces in quality but orders of magnitudes faster. Our method performs per-view dense depth optimization directly over sparse 3D Ground Control Points (GCPs), hence, removing the need for view pairing, image rectification, and stereo depth estimation, and allowing for full per-image parallelization. We use Structure-from-Motion (SfM) points as GCPs, but the method is not specific to these, e.g.~LiDAR or RGB-D can also be used. The resulting meshes are compact and inherently edge-aligned with image gradients, enabling good-quality lightweight per-face flat renderings. Our experiments demonstrate on a variety of 3D datasets the superiority in speed and competitive surface quality.

We propose an efficient and general purpose energy optimization method for binary variable energies used in various low-level vision tasks. The proposed method can be used for broad classes of higher-order and pairwise non-submodular functions. We first revisit a submodular-supermodular procedure (SSP) [Narasimhan05], which is previously studied for higher-order energy optimization. We then present our method as generalization of SSP, which is further shown to generalize several state-of-the-art techniques for higher-order and pairwise non-submodular functions [Ayed13, Gorelick14, Tang14]. In the experiments, we apply our method to image segmentation, deconvolution, and binarization, and show improvements over state-of-the-art methods.

Person re-identification has been widely studied due to its importance in surveillance and forensics applications. In practice, gallery images are high-resolution (HR) while probe images are usually low-resolution (LR) in the identification scenarios with large variation of illumination, weather or quality of cameras. Person re-identification in this kind of scenarios, which we call super-resolution (SR) person re-identification, has not been well studied. In this paper, we propose a semi-coupled low-rank discriminant dictionary learning (SLD^2L) approach for SR person re-identification. For the given training image set which consists of HR gallery and LR probe images, we aim to convert the features of LR images into discriminating HR features. Specifically, our approach learns a pair of HR and LR dictionaries and a mapping from the features of HR gallery images and LR probe images. To ensure that the converted features using the learned dictionaries and mapping have favorable discriminative capability, we design a discriminant term which requires the converted HR features of LR probe images should be close to the features of HR gallery images from the same person, but far away from the features of HR gallery images from different persons. In addition, we apply low-rank regularization in dictionary learning procedure such that the learned dictionaries can well characterize intrinsic feature space of HR and LR images. Experimental results on public datasets demonstrate the effectiveness of SLD^2L.

Subspace clustering is a problem of finding a multisubspace representation that best fits sample points drawn from a high-dimensional space. The existing clustering models generally adopt different norms to describe noise, which is equivalent to assuming that the data are corrupted by specific types of noise. In practice, however, noise is much more complex. So it is inappropriate to simply use a certain norm to model noise. Therefore, we propose Mixture of Gaussian Regression (MoG Regression) for subspace clustering by modeling noise as a Mixture of Gaussians (MoG). The MoG Regression provides an effective way to model a much broader range of noise distributions. As a result, the obtained affinity matrix is better at characterizing the structure of data in real applications. Experimental results on multiple datasets demonstrate that MoG Regression significantly outperforms state-of-the-art subspace clustering methods.

Feature correspondence plays a central role in various computer vision applications. It is widely formulated as a graph matching problem due to its robust performance under challenging conditions, such as background clutter, object deformation and repetitive patterns. A variety of fast and accurate algorithms have been proposed for graph matching. However, most of them focus on improving the recall of the solution while rarely considering its precision, thus inducing a solution with numerous outliers. To address both precision and recall feature correspondence should rather be formulated as a subgraph matching problem. This paper proposes a new subgraph matching formulation which uses a compactness prior, an additional constraint that prefers sparser solutions and effectively eliminates outliers. To solve the new optimization problem, we propose a meta-algorithm based on Markov chain Monte Carlo. By constructing Markov chain on the restricted search space instead of the original solution space, our method approximates the solution effectively. The experiments indicate that our proposed formulation and algorithm significantly improve the baseline performance under challenging conditions when both outliers and deformation noise are present.

Tracking multiple targets in a video, based on a finite set of detection hypotheses, is a persistent problem in computer vision. A common strategy for tracking is to first select hypotheses spatially and then to link these over time while maintaining disjoint path constraints. In crowded scenes multiple hypotheses will often be similar to each other making selection of optimal links an unnecessary hard optimization problem due to the sequential treatment of space and time. Embracing this observation, we propose to link and cluster plausible detections jointly across space and time. Specifically, we state multi-target tracking as a Minimum Cost Subgraph Multicut Problem. Evidence about pairs of detection hypotheses is incorporated whether the detections are in the same frame, neighboring frames or distant frames. This facilitates long-range re-identification and within-frame clustering. Results for published benchmark sequences demonstrate the superiority of this approach.

Subspace clustering refers to the problem of segmenting data drawn from a union of subspaces. State of the art approaches for solving this problem follow a two-stage approach. In the first step, an affinity matrix is learned from the data using sparse or low-rank minimization techniques. In the second step, the segmentation is found by applying spectral clustering to this affinity. While this approach has led to state of the art results in many applications, it is sub-optimal because it does not exploit the fact that the affinity and the segmentation depend on each other. In this paper, we propose a unified optimization framework for learning both the affinity and the segmentation. Our framework is based on expressing each data point as a structured sparse linear combination of all other data points, where the structure is induced by a norm that depends on the unknown segmentation. We show that both the segmentation and the structured sparse representation can be found via a combination of an alternating direction method of multipliers with spectral clustering. Experiments on a synthetic data, the Hopkins 155 motion segmentation database, and the Extended Yale B data set demonstrate the effectiveness of our approach.

Sparse representation has been applied to visual tracking by finding the best target candidate with minimal reconstruction error by use of target templates. However, most sparse representation based trackers only consider holistic or local representations and do not make full use of the intrinsic structure among and inside target candidates, thereby making the representation less effective when similar objects appear or under occlusion. In this paper, we propose a novel Structural Sparse Tracking (SST) algorithm, which not only exploits the intrinsic relationship among target candidates and their local patches to learn their sparse representations jointly, but also preserves the spatial layout structure among the local patches inside each target candidate. We show that our SST algorithm accommodates most existing sparse trackers with the respective merits. Both qualitative and quantitative evaluations on challenging benchmark image sequences demonstrate that the proposed SST algorithm performs favorably against several state-of-the-art methods.