CVPR 2015 Papers

This paper presents a novel approach to solving optical flow problems using a discrete, tree-structured MRF derived from a hierarchical segmentation of the image. Our method can be used to find globally optimal matching solutions even for problems involving very large motions. Experiments demonstrate that our approach is competitive on the MPI-Sintel dataset and that it can significantly outperform existing methods on problems involving large motions.

Pose variation remains one of the major factors adversely affect the accuracy of real-world face recognition systems. Inspired by the recently proposed probabilistic elastic part (PEP) model and the success of the deep hierarchical architecture in a number of visual tasks, we propose the Hierarchical-PEP model to approach the unconstrained face recognition problem. We apply the PEP model hierarchically to decompose a face image into face parts at different levels of details to build pose-invariant part-based face representations. Following the hierarchy from bottom-up, we stack the face part representations at each layer, discriminatively reduce its dimensionality, and hence aggregate the face part representations layer-by-layer to build a compact and invariant face representation. The Hierarchical-PEP model exploits the fine-grained structures of the face parts at different levels of details to address the pose variations. It is also guided by supervised information in constructing the face part/face representations. We empirically verify the Hierarchical-PEP model on two public benchmarks (i.e., the LFW and YouTube Faces) and a face recognition challenge (i.e., the PaSC grand challenge) for image-based and video-based face verification. The state-of-the-art performance demonstrates the potential of our method.

We address the problem of estimating location information of an image using principles from automated representation learning. We pursue a hierarchical sparse coding approach that learns features useful in discriminating images across locations, by initializing it with a geometric prior corresponding to transformations between image appearance space and their corresponding location grouping space using the notion of parallel transport on manifolds. We then extend this approach to account for the availability of heterogeneous data modalities such as geo-tags and videos pertaining to different locations, and also study a relatively under-addressed problem of transferring knowledge available from certain locations to infer the grouping of data from novel locations. We evaluate our approach on several standard datasets such as im2gps, San Francisco and MediaEval2010, and obtain state-of-the-art results.

Human actions can be represented by the trajectories of skeleton joints. Traditional methods generally model the spatial structure and temporal dynamics of human skeleton with hand-crafted features and recognize human actions by well-designed classifiers. In this paper, considering that recurrent neural network (RNN) can model the long-term contextual information of temporal sequences well, we propose an end-to-end hierarchical RNN for skeleton based action recognition. Instead of taking the whole skeleton as the input, we divide the human skeleton into five parts according to human physical structure, and then separately feed them to five subnets. As the number of layers increases, the representations extracted by the subnets are hierarchically fused to be the inputs of higher layers. The final representations of the skeleton sequences are fed into a single-layer perceptron, and the temporally accumulated output of the perceptron is the final decision. We compare with five other deep RNN architectures derived from our model to verify the effectiveness of the proposed network, and also compare with several other methods on three publicly available datasets. Experimental results demonstrate that our model achieves the state-of-the-art performance with high computational efficiency.

Many discriminant analysis methods such as LDA and HLDA actually maximize the average pairwise distances between classes, which often causes the class separation problem. Max-min distance analysis (MMDA) addresses this problem by maximizing the minimum pairwise distance in the latent subspace, but it is developed under the homoscedastic assumption. This paper proposes Heteroscedastic MMDA (HMMDA) methods that explore the discriminative information in the difference of intra-class scatters for dimensionality reduction. WHMMDA maximizes the minimal pairwise Chenoff distance in the whitened space. OHMMDA incorporates this objective and the minimization of class compactness into a trace quotient formulation and imposes an orthogonal constraint to the final transformation, which can be solved by a bisection search algorithm. Two variants of OHMMDA are further proposed to encode the margin information. Experiments on several UCI Machine Learning datasets and the Yale Face database demonstrate the effectiveness of the proposed HMMDA methods.

This paper propose an approach for modeling textured 3D non-rigid models based on Weighted Heat Kernel Signature(W-HKS). As a first contribution, we show how to include photometric information as a weight over the shape manifold, we also propose a novel formulation for heat diffusion over weighted manifolds. As a second contribution we present a new discretization method for the proposed equation using finite element approximation. Finally, the weighted heat kernel signature is used as a shape descriptor. The proposed descriptor encodes both the photometric, and geometric information based on the solution of one equation. We also propose a new method to introduce the scale invariance for the weighted heat kernel signature. The performance is tested on two benchmark datasets. The results have indeed confirmed the high performance of the proposed approach on the textured shape retrieval problem, and show that the proposed method is useful in coping with different challenges of shape analysis where pure geometric and pure photometric methods fail.

An attractive approach for fast search in image databases is binary hashing, where each high-dimensional, real-valued image is mapped onto a low-dimensional, binary vector and the search is done in this binary space. Finding the optimal hash function is difficult because it involves binary constraints, and most approaches approximate the optimization by relaxing the constraints and then binarizing the result. Here, we focus on the binary autoencoder model, which seeks to reconstruct an image from the binary code produced by the hash function. We show that the optimization can be simplified with the method of auxiliary coordinates. This reformulates the optimization as alternating two easier steps: one that learns the encoder and decoder separately, and one that optimizes the code for each image. Image retrieval experiments show the resulting hash function outperforms or is competitive with state-of-the-art methods for binary hashing.

In recent years, several feature encoding schemes for the bags-of-visual-words model have been proposed. While most of these schemes produce impressive results, they all share an important limitation: their high computational complexity makes it challenging to use them for large-scale problems. In this work, we propose an approximate locality-constrained encoding scheme that offers significantly better computational efficiency (~40x) than its exact counterpart, with comparable classification accuracy. Using the perturbation analysis of least-squares problems, we present a formal approximation error analysis of our approach, which helps distill the intuition behind the robustness of our method. We present a thorough set of empirical analyses on multiple standard data-sets, to assess the capability of our encoding scheme for its representational as well as discriminative accuracy.

Ubiquitous motion blur easily fails multi-frame super-resolution (MFSR). Our method proposed in this paper tackles this issue by optimally searching least blurred pixels in MFSR. An EM framework is proposed to guide residual blur estimation and high-resolution image reconstruction. To suppress noise, we employ a family of sparse penalties as natural image priors, along with an effective solver. Theoretical analysis is performed on how and when our method works. The relationship between estimation errors of motion blur and the quality of input images is discussed. Our method produces sharp and higher-resolution results given input of challenging low-resolution noisy and blurred sequences.

The mainstream approach to structured prediction problems in computer vision is to learn an energy function such that the solution minimizes that function. At prediction time, this approach must solve an often-challenging optimization problem. Search-based methods provide an alternative that has the potential to achieve higher performance. These methods learn to control a search procedure that constructs and evaluates candidate solutions. The recently-developed HC-Search method has been shown to achieve state-of-the-art results in natural language processing, but mixed success when applied to vision problems. This paper studies whether HC-Search can achieve similarly competitive performance on basic vision tasks such as object detection, scene labeling, and monocular depth estimation, where the leading paradigm is energy minimization. To this end, we introduce a search operator suited to the vision domain that improves a candidate solution by probabilistically sampling likely object configurations in the scene from the hierarchical Berkeley segmentation. We complement this search operator by applying the DAgger algorithm to robustly train the search heuristic so it learns from its previous mistakes. Our evaluation shows that these improvements reduce the branching factor and search depth, and thus give a significant performance boost. Our state-of-the-art results on scene labeling and depth estimation suggest that HC-Search provides a suitable tool for learning and inference in vision.

The grasp type provides crucial information about human action. However, recognizing the grasp type in unconstrained scenes is challenging because of the large variations in appearance, occlusions and geometric distortions. In this paper, first we present a convolutional neural network to classify functional hand grasp types. Experiments on a public static scene hand data set validate good performance of the presented method. Then we present two applications utilizing grasp type classification: (a) inference of human action intention and (b) fine level manipulation action segmentation. Experiments on both tasks demonstrate the usefulness of grasp type as a cognitive feature for computer vision. This study shows that the grasp type is a powerful symbolic representation for action understanding, and thus opens new avenues for future research.

We show how the simplex algorithm can be tailored to the linear programming relaxation of pairwise energy minimization with binary variables. A special structure formed by basic and nonbasic variables in each stage of the algorithm is identified and utilized to perform the whole iterative process combinatorially over the input energy minimization graph rather than algebraically over the simplex tableau. This leads to a new efficient solver. We demonstrate that for some computer vision instances it performs even better than methods reducing binary energy minimization to finding maximum flow in a network.

Not all measured features in SLAM/SfM contribute to accurate localization during the estimation process, thus it is sensible to utilize only those that do. This paper describes a method for selecting a subset of features that are of high utility for localization in the SLAM/SfM estimation process. It is derived by examining the observability of SLAM and, being complimentary to the estimation process, it easily integrates into existing SLAM systems. The measure of estimation utility is formulated with temporal and instantaneous observability indices. Efficient computation strategies for the observability indices are described based on incremental singular value decomposition (SVD) and greedy selection for the temporal and instantaneous observability indices, respectively. The greedy selection is near-optimal since the observability index is (approximately) submodular. The proposed method improves localization and data association. Controlled synthetic experiments with ground truth demonstrate the improved localization accuracy, and real-time SLAM experiments demonstrate the improved data association.

We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC2014). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation of this architecture, GoogLeNet, a 22 layers deep network, was used to assess its quality in the context of object detection and classification.

Mathematical optimization plays a fundamental role in solving many problems in computer vision (e.g., camera calibration, image alignment, structure from motion). It is generally accepted that second order descent methods are the most robust, fast, and reliable approaches for nonlinear optimization of a general smooth function. However, in the context of computer vision, second order descent methods have two main drawbacks: 1) the function might not be analytically differentiable and numerical approximations are impractical, and 2) the Hessian may be large and not positive definite. Recently, Supervised Descent Method (SDM), a method that learns the "weighted averaged gradients" in a supervised manner has been proposed to solve these issues. However, SDM is a local algorithm and it is likely to average conflicting gradient directions. This paper proposes Global SDM (GSDM), an extension of SDM that divides the search space into regions of similar gradient directions. GSDM provides a better and more efficient strategy to minimize non-linear least squares functions in computer vision. We illustrate the effectiveness of GSDM in two problems: non-rigid image alignment and extrinsic camera calibration.

Random forest is well known as one of the best learning methods. In spite of its great success, it also has certain drawbacks: the heuristic learning rule does not effectively minimize the global training loss; the model size is usually too large for many real applications. To address the issues, we propose two techniques, global refinement and global pruning, to improve a pre-trained random forest. The proposed global refinement jointly relearns the leaf nodes of all trees under a global objective function so that the complementary information between multiple trees is well exploited. In this way, the fitting power of the forest is significantly enhanced. The global pruning is developed to reduce the model size as well as the over-fitting risk. The refined model has better performance and smaller storage cost, as verified in extensive experiments.

We consider kernel methods on general geodesic metric spaces and provide both negative and positive results. First we show that the common Gaussian kernel can only be generalized to a positive definite kernel on a geodesic metric space if the space is flat. As a result, for data on a Riemannian manifold, the geodesic Gaussian kernel is only positive definite if the Riemannian manifold is Euclidean. This implies that any attempt to design geodesic Gaussian kernels on curved Riemannian manifolds is futile. However, we show that for spaces with conditionally negative definite distances the geodesic Laplacian kernel can be generalized while retaining positive definiteness. This implies that geodesic Laplacian kernels can be generalized to some curved spaces, including spheres and hyperbolic spaces. Our theoretical results are verified empirically.

The availability of GIS (Geographical Information System) databases for many urban areas, provides a valuable source of information for improving the performance of many computer vision tasks. In this paper, we propose a method which leverages information acquired from GIS databases to perform semantic segmentation of the image alongside with geo-referencing each semantic segment with its address and geo-location. First, the image is segmented into a set of initial super-pixels. Then, by projecting the information from GIS databases, a set of priors are obtained about the approximate location of the semantic entities such as buildings and streets in the image plane. However, there are significant inaccuracies (misalignments) in the projections, mainly due to inaccurate GPS-tags and camera parameters. In order to address this misalignment issue, we perform data fusion such that it improves the segmentation and GIS projections accuracy simultaneously with an iterative approach. At each iteration, the projections are evaluated and weighted in terms of reliability, and then fused with the super-pixel segmentations. First segmentation is performed using random walks, based on the GIS projections. Then the global transformation which best aligns the projections to their corresponding semantic entities is computed and applied to the projections to further align them to the content of the image. The iterative approach continues until the projections and segments are well aligned.

Several state-of-the-art video deblurring methods are based on a strong assumption that the captured scenes are static. These methods fail to deblur blurry videos in dynamic scenes. We propose a video deblurring method to deal with general blurs inherent in dynamic scenes, contrary to other methods. To handle locally varying and general blurs caused by various sources, such as camera shake, moving objects, and depth variation in a scene, we approximate pixel-wise kernel with bidirectional optical flows. Therefore, we propose a single energy model that simultaneously estimates optical flows and latent frames to solve our deblurring problem. We also provide a framework and efficient solvers to optimize the energy model. By minimizing the proposed energy function, we achieve significant improvements in removing blurs and estimating accurate optical flows in blurry frames. Extensive experimental results demonstrate the superiority of the proposed method in real and challenging videos that state-of-the-art methods fail in either deblurring or optical flow estimation.