CVPR 2016 Papers

Previous work on estimating the epipolar geometry of two views relies on being able to reliably match feature points based on appearance. In this paper, we go one step further and show that it is feasible to compute both the epipolar geometry and the correspondences at the same time based on geometry only. We do this in a globally optimal manner. Our approach is based on an efficient branch and bound technique in combination with bipartite matching to solve the correspondence problem. We rely on several recent works to obtain good bounding functions to battle the combinatorial explosion of possible matchings. It is experimentally demonstrated that more difficult cases can be handled and that more inlier correspondences can be obtained by being less restrictive in the matching phase.

Existing optical flow methods make generic, spatially homogeneous, assumptions about the spatial structure of the flow. In reality, optical flow varies across an image depending on object class. Simply put, different objects move differently. Here we exploit recent advances in static semantic scene segmentation to segment the image into objects of different types. We define different models of image motion in these regions depending on the type of object. For example, the road motion with homographies, vegetation with spatially smooth flow, and independently moving objects like cars and planes with affine+deviations. We then pose the flow estimation problem using a novel formulation of localized layers, which addresses limitations of traditional layered models for dealing with complex scene motion. Our semantic flow method achieves the lowest error of any published method in the KITTI-2015 flow benchmark and produces qualitatively better flow and segmentation than recent top methods on a wide range of natural videos.

A system capable of performing robust online volumetric reconstruction of indoor scenes based on input from a handheld RGB-D camera is presented. Our system is powered by a two-pass reconstruction scheme. The first pass tracks camera poses at video rate and simultaneously constructs a pose graph on-the-fly. The tracker operates in real-time, which allows the reconstruction results to be visualized during the scanning process. Live visual feedbacks makes the scanning operation fast and intuitive. Upon termination of scanning, the second pass takes place to handle loop closures and reconstruct the final model using globally refined camera trajectories. The system is online with low delay and returns a dense model of sufficient accuracy. The beauty of this system lies in its speed, accuracy, simplicity and ease of implementation when compared to previous methods. We demonstrate the performance of our system on several real-world scenes and quantitatively assess the modeling accuracy with respect to ground truth models obtained from a LIDAR scanner.

Multi-object tracking (MOT) becomes more challenging when objects of interest have similar appearances. In that case, the motion cues are particularly useful for discriminating multiple objects. However, for online 2D MOT in scenes acquired from moving cameras, observable motion cues are complicated by global camera movements and thus not always smooth or predictable. To deal with such unexpected camera motion for online 2D MOT, a structural motion constraint between objects has been utilized thanks to its robustness to camera motion. In this paper, we propose a new data association method that effectively exploits structural motion constraints in the presence of large camera motion. In addition, to further improve the robustness of data association against mis-detections and clutters, a novel event aggregation approach is developed to integrate structural constraints in assignment costs for online MOT. Experimental results on a large number of datasets demonstrate the effectiveness of the proposed algorithm for online 2D MOT.

Randomized classification trees are among the most popular machine learning tools and found successful applications in many areas. Although this classifier was originally designed as offline learning algorithm, there has been an increased interest in the last years to provide an online variant. In this paper, we propose an online learning algorithm for classification trees that adheres to Bayesian principles. In contrast to state-of-the-art approaches that produce large forests with complex trees, we aim at constructing small ensembles consisting of shallow trees with high generalization capabilities. Experiments on benchmark machine learning and body part recognition datasets show superior performance over state-of-the-art approaches.

Automatic detection and classification of dynamic hand gestures in real-world systems intended for human computer interaction is challenging as: 1) there is a large diversity in how people perform gestures, making detection and classification difficult; 2) the system must work online in order to avoid noticeable lag between performing a gesture and its classification; in fact, a negative lag (classification before the gesture is finished) is desirable, as feedback to the user can then be truly instantaneous. In this paper, we address these challenges with a recurrent three-dimensional convolutional neural network that performs simultaneous detection and classification of dynamic hand gestures from multi-modal data. We employ connectionist temporal classification to train the network to predict class labels from in-progress gestures in unsegmented input streams. In order to validate our method, we introduce a new challenging multi-modal dynamic hand gesture dataset captured with depth, color and stereo-IR sensors. On this challenging dataset, our gesture recognition system achieves an accuracy of 83.8%, outperforms competing state-of-the-art algorithms, and approaches human accuracy of 88.4%. Moreover, our method achieves state-of-the-art performance on SKIG and ChaLearn2014 benchmarks.

We focus on learning open-vocabulary visual classifiers, which scale up to a large portion of natural language vocabulary (e.g., over tens of thousands of classes). In particular, the training data are large-scale weakly labeled Web images since it is difficult to acquire sufficient well-labeled data at this category scale. In this paper, we propose a novel online learning paradigm towards this challenging task. Different from traditional N-way independent classifiers that generally fail to handle the extremely sparse and inter-related labels, our classifiers learn from continuous label embeddings discovered by collaboratively decomposing the sparse image-label matrix. Leveraging on the structure of the proposed collaborative learning formulation, we develop an efficient online algorithm that can jointly learn the label embeddings and visual classifiers. The algorithm can learn over 30,000 classes of 1,000 training images within 1 second on a standard GPU. Extensively experimental results on four benchmarks demonstrate the effectiveness of our method.

The appearance of (outdoor) scenes changes considerably with the strength of certain transient attributes, such as "rainy", "dark" or "sunny". Obviously, this also affects the representation of an image in feature space, e.g., as activations at a certain CNN layer, and consequently impacts scene recognition performance. In this work, we investigate the variability in these transient attributes as a rich source of information for studying how image representations change as a function of attribute strength. In particular, we leverage a recently introduced dataset with fine-grain annotations to estimate feature trajectories for a collection of transient attributes and then show how these trajectories can be transferred to new image representations. This enables us to synthesize new data along the transferred trajectories with respect to the dimensions of the space spanned by the transient attributes. Applicability of this concept is demonstrated on the problem of one-shot recognition of scene locations. We show that data synthesized via feature trajectory transfer considerably boosts recognition performance, (1) with respect to baselines and (2) in combination with state-of-the-art approaches in one-shot learning.

Sparse feature (dictionary) selection is critical for various tasks in computer vision, machine learning, and pattern recognition to avoid overfitting. While extensive research efforts have been conducted on feature selection using sparsity and group sparsity, we note that there has been a lack of development on applications where there is a particular preference on diversity. That is, the selected features are expected to come from different groups or categories. This diversity preference is motivated from many real-world applications such as advertisement recommendation, privacy image classification, and design of survey. In this paper, we proposed a general bilevel exclusive sparsity formulation to pursue the diversity by restricting the overall sparsity and the sparsity in each group. To solve the proposed formulation that is NP hard in general, a heuristic procedure is proposed. The main contributions in this paper include: 1) A linear convergence rate is established for the proposed algorithm; 2) The provided theoretical error bound improves the approaches such as L_1 norm and L_0 types methods which only use the overall sparsity and the quantitative benefits of using the diversity sparsity is provided. To the best of our knowledge, this is the first work to show the theoretical benefits of using the diversity sparsity; 3) Extensive empirical studies are provided to validate the proposed formulation, algorithm, and theory.

Face alignment or facial landmark detection plays an important role in many computer vision applications, e.g., face recognition, facial expression recognition, face animation, etc. However, the performance of face alignment system degenerates severely when occlusions occur. In this work, we propose a novel face alignment method, which cascades several Deep Regression networks coupled with De-corrupt Autoencoders (denoted as DRDA) to explicitly handle partial occlusion problem. Different from the previous works that can only detect occlusions and discard the occluded parts, our proposed de-corrupt autoencoder network can automatically recover the genuine appearance for the occluded parts and the recovered parts can be leveraged together with those non-occluded parts for more accurate alignment. By coupling de-corrupt autoencoders with deep regression networks, a deep alignment model robust to partial occlusions is achieved. Besides, our method can localize occluded regions rather than merely predict whether the landmarks are occluded. Experiments on two challenging occluded face datasets demonstrate that our method significantly outperforms the state-of-the-art methods.

Occlusion boundaries contain rich perceptual information about the underlying scene structure. They also provide important cues in many visual perception tasks such as scene understanding, object recognition, and segmentation. In this paper, we improve occlusion boundary detection via enhanced exploration of contextual information (e.g., local structural boundary patterns, observations from surrounding regions, and temporal context), and in doing so develop a novel approach based on convolutional neural networks (CNNs) and conditional random fields (CRFs). Experimental results demonstrate that our detector significantly outperforms the state-of-the-art (e.g., improving the F-measure from 0.62 to 0.71 on the commonly used CMU benchmark). Last but not least, we empirically assess the roles of several important components of the proposed detector, so as to validate the rationale behind this approach.

Object proposals have quickly become the de-facto pre-processing step in a number of vision pipelines (for object detection, object discovery, and other tasks). Their performance is usually evaluated on partially annotated datasets. In this paper, we argue that the choice of using a partially annotated dataset for evaluation of object proposals is problematic -- as we demonstrate via a thought experiment, the evaluation protocol is 'gameable', in the sense that progress under this protocol does not necessarily correspond to a "better" category independent object proposal algorithm. To alleviate this problem, we: (1) Introduce a nearly-fully annotated version of PASCAL VOC dataset, which serves as a test-bed to check if object proposal techniques are overfitting to a particular list of categories. (2) Perform an exhaustive evaluation of object proposal methods on our introduced nearly-fully annotated PASCAL dataset and perform cross-dataset generalization experiments; and (3) Introduce a diagnostic experiment to detect the bias capacity in an object proposal algorithm. This tool circumvents the need to collect a densely annotated dataset, which can be expensive and cumbersome to collect. Finally, we have released an easy-to-use toolbox which combines various publicly available implementations of object proposal algorithms which standardizes the proposal generation and evaluation so that new methods can be added and evaluated on different datasets. We hope that the results presented in the paper will motivate the community to test the category independence of various object proposal methods by carefully choosing the evaluation protocol.

Structured support vector machine (SSVM) based methods has demonstrated encouraging performance in recent object tracking benchmarks. However, the complex and expensive optimization limits their deployment in real-world applications. In this paper, we present a simple yet efficient dual linear SSVM (DLSSVM) algorithm to enable fast learning and execution during tracking. By analyzing the dual variables, we propose a primal classifier update formula where the learning step size is computed in closed form. This online learning method significantly improves the robustness of the proposed linear SSVM with low computational cost. Second, we approximate the intersection kernel for feature representations with an explicit feature map to further improve tracking performance. Finally, we extend the proposed DLSSVM tracker in a multiscale manner to address the "drift" problem. Experimental results on large benchmark datasets with 50 and 100 video sequences show that the proposed DLSSVM tracking algorithm achieves state-of-the-art performance.

Object skeleton is a useful cue for object detection, complementary to the object contour, as it provides a structural representation to describe the relationship among object parts. While object skeleton extraction in natural images is a very challenging problem, as it requires the extractor to be able to capture both local and global image context to determine the intrinsic scale of each skeleton pixel. Existing methods rely on per-pixel based multi-scale feature computation, which results in difficult modeling and high time consumption. In this paper, we present a fully convolutional network with multiple scale-associated side outputs to address this problem. By observing the relationship between the receptive field sizes of the sequential stages in the network and the skeleton scales they can capture, we introduce a scale-associated side output to each stage. We impose supervision to different stages by guiding the scale-associated side outputs toward groundtruth skeletons of different scales. The responses of the multiple scale-associated side outputs are then fused in a scale-specific way to localize skeleton pixels with multiple scales effectively. Our method achieves promising results on two skeleton extraction datasets, and significantly outperforms other competitors.

Deep Convolution Neural Networks (CNNs) have shown impressive performance in various vision tasks such as image classification, object detection and semantic segmentation. For object detection, particularly in still images, the performance has been significantly increased last year thanks to powerful deep networks (e.g. GoogleNet) and detection frameworks (e.g. Regions with CNN features (R-CNN)). The lately introduced ImageNet task on object detection from video (VID) brings the object detection task into video domain, in which objects' locations at each frame are required to be annotated with bounding boxes. In this work, we introduce a complete framework for the VID task based on still-image object detection and general object tracking. Their relations and contributions in the VID task are thoroughly studied and evaluated. In addition, a temporal convolution network is proposed to incorporate temporal information to regularize the detection results and shows its effectiveness for the task.

We develop a deep learning algorithm for contour detection with a fully convolutional encoder-decoder network. Different from previous low-level edge detection, our algorithm focuses on detecting higher-level object contours. Our network is trained end-to-end on PASCAL VOC with refined ground truth from inaccurate polygon annotations, yielding much higher precision in object contour detection than previous methods. We find that the learned model generalizes well to unseen object classes from the same supercategories on MS COCO and can match state-of-the-art edge detection on BSDS500 with fine-tuning. By combining with the multiscale combinatorial grouping algorithm, our method can generate high-quality segmented object proposals, which significantly advance the state-of-the-art on PASCAL VOC (improving average recall from 0.62 to 0.67) with a relatively small amount of candidates (1660 per image).

Aiming at automatically discovering the common objects contained in a set of relevant images and segmenting them as foreground simultaneously, object co-segmentation has become an active research topic in recent years. Although a number of approaches have been proposed to address this problem, many of them are designed with the misleading assumption, unscalable prior, or low flexibility and thus still suffer from certain limitations, which reduces their capability in the real-world scenarios. To alleviate these limitations, we propose a novel two-stage co-segmentation framework, which introduces the weak background prior to establish a globally close- loop graph to represent the common object and union background separately. Then a novel graph optimized-flexible manifold ranking algorithm is proposed to flexibly optimize the graph connection and node labels to co-segment the common objects. Experiments on three image datasets demonstrate that our method outperforms other state-of-the-art methods.

Haze limits visibility and reduces image contrast in outdoor images. The degradation is different for every pixel and depends on the distance of the scene point from the camera. This dependency is expressed in the transmission coefficients, that control the scene attenuation and amount of haze in every pixel. Previous methods solve the single image dehazing problem using various patch-based priors. We, on the other hand, propose an algorithm based on a new, non-local prior. The algorithm relies on the assumption that colors of a haze-free image are well approximated by a few hundred distinct colors, that form tight clusters in RGB space. Our key observation is that pixels in a given cluster are often non-local, i.e., they are spread over the entire image plane and are located at different distances from the camera. In the presence of haze these varying distances translate to different transmission coefficients. Therefore, each color cluster in the clear image becomes a line in RGB space, that we term a haze-line. Using these haze-lines, our algorithm recovers both the distance map and the haze-free image. The algorithm is linear in the size of the image, deterministic and requires no training. It performs well on a wide variety of images and is competitive with other state-of-the-art methods.

Recent shadow detection algorithms have shown initial success on small datasets of images from specific domains. However, shadow detection on broader image domains is still challenging due to the lack of annotated training data. This is due to the intense manual labor in annotating shadow data. In this paper we propose "lazy annotation", an efficient annotation method where an annotator only needs to mark the important shadow areas and some non-shadow areas. This yields data with noisy labels that are not yet useful for training a shadow detector. We address the problem of label noise by jointly learning a shadow region classifier and recovering the labels in the training set. We consider the training labels as unknowns and formulate the label recovery problem as the minimization of the sum of squared leave-one-out errors of a Least Squares SVM, which can be efficiently optimized. Experimental results show that a classifier trained with recovered labels achieves comparable performance to a classifier trained on the properly annotated data. These results suggest a feasible approach to address the task of detecting shadows in an unfamiliar domain: collecting and lazily annotating some images from the new domain for training. As will be demonstrated, this approach outperforms methods that rely on precisely annotated but less relevant datasets. Initial results suggest more general applicability.