ICCV 2015 论文列表

Active learning is an effective way to relieve the tedious work of manual annotation in many applications of visual recognition. However, less research attention has been focused on multi-class active learning. In this paper, we propose a novel Gaussian process classifier model with multiple annotators for multi-class visual recognition. Expectation propagation (EP) is adopted for efficient approximate Bayesian inference of our probabilistic model for classification. Based on the EP approximation inference, a generalized Expectation Maximization (GEM) algorithm is derived to estimate both the parameters for instances and the quality of each individual annotator. Also, we incorporate the idea of reinforcement learning to actively select both the informative samples and the high-quality annotators, which better explores the trade-off between exploitation and exploration. The experiments clearly demonstrate the efficacy of the proposed model.

For the segmentation of moving objects in videos, the analysis of long-term point trajectories has been very popular recently. In this paper, we formulate the segmentation of a video sequence based on point trajectories as a minimum cost multicut problem. Unlike the commonly used spectral clustering formulation, the minimum cost multicut formulation gives natural rise to optimize not only for a cluster assignment but also for the number of clusters while allowing for varying cluster sizes. In this setup, we provide a method to create a long-term point trajectory graph with attractive and repulsive binary terms and outperform state-of-the-art methods based on spectral clustering on the FBMS-59 dataset and on the motion subtask of the VSB100 dataset.

In this paper we tackle the problem of instance-level segmentation and depth ordering from a single monocular image. Towards this goal, we take advantage of convolutional neural nets and train them to directly predict instance-level segmentations where the instance ID encodes the depth ordering within image patches. To provide a coherent single explanation of an image we develop a Markov random field which takes as input the predictions of convolutional neural nets applied at overlapping patches of different resolutions, as well as the output of a connected component algorithm. It aims to predict accurate instance-level segmentation and depth ordering. We demonstrate the effectiveness of our approach on the challenging KITTI benchmark and show good performance on both tasks.

Consumer depth cameras have dramatically improved our ability to track rigid, articulated, and deformable 3D objects in real-time. However, depth cameras have a limited temporal resolution (frame-rate) that restricts the accuracy and robustness of tracking, especially for fast or unpredictable motion. In this paper, we show how to perform model-based object tracking which allows to reconstruct the object's depth at an order of magnitude higher frame-rate through simple modifications to an off-the-shelf depth camera. We focus on phase-based time-of-flight (ToF) sensing, which reconstructs each low frame-rate depth image from a set of short exposure 'raw' infrared captures. These raw captures are taken in quick succession near the beginning of each depth frame, and differ in the modulation of their active illumination. We make two contributions. First, we detail how to perform model-based tracking against these raw captures. Second, we show that by reprogramming the camera to space the raw captures uniformly in time, we obtain a 10x higher frame-rate, and thereby improve the ability to track fast-moving objects.

In this paper, we propose a novel geometric model fitting method, called Mode-Seeking on Hypergraphs (MSH), to deal with multi-structure data even in the presence of severe outliers. The proposed method formulates geometric model fitting as a mode seeking problem on a hypergraph in which vertices represent model hypotheses and hyperedges denote data points. MSH intuitively detects model instances by a simple and effective mode seeking algorithm. In addition to the mode seeking algorithm, MSH includes a similarity measure between vertices on the hypergraph and a "weight-aware sampling" technique. The proposed method not only alleviates sensitivity to the data distribution, but also is scalable to large scale problems. Experimental results further demonstrate that the proposed method has significant superiority over the state-of-the-art fitting methods on both synthetic data and real images.

This paper reformulates the theory of graph mining on the technical basis of graph matching, and extends its scope of applications to computer vision. Given a set of attributed relational graphs (ARGs), we propose to use a hierarchical And-Or Graph (AoG) to model the pattern of maximal-size common subgraphs embedded in the ARGs, and we develop a general method to mine the AoG model from the unlabeled ARGs. This method provides a general solution to the problem of mining hierarchical models from unannotated visual data without exhaustive search of objects. We apply our method to RGB/RGB-D images and videos to demonstrate its generality and the wide range of applicability. The code will be available at https://sites.google.com/site/quanshizhang/mining-and-or-graphs.

We address the problem of minimizing human effort in interactive tracking by learning sequence-specific model parameters. Determining the optimal model parameters for each sequence is a critical problem in tracking. We demonstrate that by using the optimal model parameters for each sequence we can achieve high precision tracking results with significantly less effort. We leverage the sequential nature of interactive tracking to formulate an efficient method for learning model parameters through a maximum margin framework. By using our method we are able to save 60-90% of human effort to achieve high precision on two datasets: the VIRAT dataset and an Infant-Mother Interaction dataset.

We propose two novel minimal solvers which advance the state of the art in satellite imagery processing. Our methods are efficient and do not rely on the prior existence of complex inverse mapping functions to correlate 2D image coordinates and 3D terrain. Our first solver improves on the stereo correspondence problem for satellite imagery, in that we provide an exact image-to-object space mapping (where prior methods were inaccurate). Our second solver provides a novel mechanism for 3D point triangulation, which has improved robustness and accuracy over prior techniques. Given the usefulness and ubiquity of satellite imagery, our proposed methods allow for improved results in a variety of existing and future applications.

Recent advances in Structure-from-Motion not only enable the reconstruction of large scale scenes, but are also able to detect ambiguous structures caused by repeating elements that might result in incorrect reconstructions. Yet, it is not always possible to fully reconstruct a scene. The images required to merge different sub-models might be missing or it might be impossible to acquire such images in the first place due to occlusions or the structure of the scene. The problem of aligning multiple reconstructions that do not have visual overlap is impossible to solve in general. An important variant of this problem is the case in which individual sides of a building can be reconstructed but not joined due to the missing visual overlap. In this paper, we present a combinatorial approach for solving this variant by automatically stitching multiple sides of a building together. Our approach exploits symmetries and semantic information to reason about the possible geometric relations between the individual models. We show that our approach is able to reconstruct complete building models where traditional SfM ends up with disconnected building sides.

This paper focuses on structured-output learning using deep neural networks for 3D human pose estimation from monocular images. Our network takes an image and 3D pose as inputs and outputs a score value, which is high when the image-pose pair matches and low otherwise. The network structure consists of a convolutional neural network for image feature extraction, followed by two sub-networks for transforming the image features and pose into a joint embedding. The score function is then the dot-product between the image and pose embeddings. The image-pose embedding and score function are jointly trained using a maximum-margin cost function. Our proposed framework can be interpreted as a special form of structured support vector machines where the joint feature space is discriminatively learned using deep neural networks. We test our framework on the Human3.6m dataset and obtain state-of-the-art results compared to other recent methods. Finally, we present visualizations of the image-pose embedding space, demonstrating the network has learned a high-level embedding of body-orientation and pose-configuration.

Deep neural network architectures have recently produced excellent results in a variety of areas in artificial intelligence and visual recognition, well surpassing traditional shallow architectures trained using hand-designed features. The power of deep networks stems both from their ability to perform local computations followed by pointwise non-linearities over increasingly larger receptive fields, and from the simplicity and scalability of the gradient-descent training procedure based on backpropagation. An open problem is the inclusion of layers that perform global, structured matrix computations like segmentation (e.g. normalized cuts) or higher-order pooling (e.g. log-tangent space metrics defined over the manifold of symmetric positive definite matrices) while preserving the validity and efficiency of an end-to-end deep training framework. In this paper we propose a sound mathematical apparatus to formally integrate global structured computation into deep computation architectures. At the heart of our methodology is the development of the theory and practice of backpropagation that generalizes to the calculus of adjoint matrix variations. We perform segmentation experiments using the BSDS and MSCOCO benchmarks and demonstrate that deep networks relying on second-order pooling and normalized cuts layers, trained end-to-end using matrix backpropagation, outperform counterparts that do not take advantage of such global layers.

We present a new, massively parallel method for high-quality multiview matching. Our work builds on the Patchmatch idea: starting from randomly generated 3D planes in scene space, the best-fitting planes are iteratively propagated and refined to obtain a 3D depth and normal field per view, such that a robust photo-consistency measure over all images is maximized. Our main novelties are on the one hand to formulate Patchmatch in scene space, which makes it possible to aggregate image similarity across multiple views and obtain more accurate depth maps. And on the other hand a modified, diffusion-like propagation scheme that can be massively parallelized and delivers dense multiview correspondence over ten 1.9-Megapixel images in 3 seconds, on a consumer-grade GPU. Our method uses a slanted support window and thus has no fronto-parallel bias; it is completely local and parallel, such that computation time scales linearly with image size, and inversely proportional to the number of parallel threads. Furthermore, it has low memory footprint (four values per pixel, independent of the depth range). It therefore scales exceptionally well and can handle multiple large images at high depth resolution. Experiments on the DTU and Middlebury multiview datasets as well as oblique aerial images show that our method achieves very competitive results with high accuracy and completeness, across a range of different scenarios.

Most of the feature-learning methods for RGB-D object recognition either learn features from color and depth modalities separately, or simply treat RGB-D as undifferentiated four-channel data, which cannot adequately exploit the relationship between different modalities. Motivated by the intuition that different modalities should contain not only some modal-specific patterns but also some shared common patterns, we propose a multi-modal feature learning framework for RGB-D object recognition. We first construct deep CNN layers for color and depth separately, and then connect them with our carefully designed multi-modal layers, which fuse color and depth information by enforcing a common part to be shared by features of different modalities. In this way, we obtain features reflecting shared properties as well as modal-specific properties in different modalities. The information of the multi-modal learning frameworks is back-propagated to the early CNN layers. Experimental results show that our proposed multi-modal feature learning method outperforms state-of-the-art approaches on two widely used RGB-D object benchmark datasets.

This work focuses on the problem of multi-label learning with missing labels (MLML), which aims to label each test instance with multiple class labels given training instances that have an incomplete/partial set of these labels (i.e. some of their labels are missing). To handle missing labels, we propose a unified model of label dependencies by constructing a mixed graph, which jointly incorporates (i) instance-level similarity and class co-occurrence as undirected edges and (ii) semantic label hierarchy as directed edges. Unlike most MLML methods, We formulate this learning problem transductively as a convex quadratic matrix optimization problem that encourages training label consistency and encodes both types of label dependencies (i.e. undirected and directed edges) using quadratic terms and hard linear constraints. The alternating direction method of multipliers (ADMM) can be used to exactly and efficiently solve this problem. To evaluate our proposed method, we consider two popular applications (image and video annotation), where the label hierarchy can be derived from Wordnet. Experimental results show that our method achieves a significant improvement over state-of-the-art methods in performance and robustness to missing labels.

A new method is introduced for stereo matching that operates on minimum spanning trees (MSTs) generated from the images. Disparity maps are represented as a collection of hidden states on MSTs, and each MST is modeled as a hidden Markov tree. An efficient recursive message-passing scheme designed to operate on hidden Markov trees, known as the upward-downward algorithm, is used to compute the maximum a posteriori (MAP) disparity estimate at each pixel. The messages processed by the upward-downward algorithm involve two types of probabilities: the probability of a pixel having a particular disparity given a set of per-pixel matching costs, and the probability of a disparity transition between a pair of connected pixels given their similarity. The distributions of these probabilities are modeled from a collection of images with ground truth disparities. Performance evaluation using the Middlebury stereo benchmark version 3 demonstrates that the proposed method ranks second and third in terms of overall accuracy when evaluated on the training and test image sets, respectively.

In this work, we propose a novel Weakly Supervised Learning (WSL) framework dedicated to learn discriminative part detectors from images annotated with a global label. Our WSL method encompasses three main contributions. Firstly, we introduce a new structured output latent variable model, Minimum mAximum lateNt sTRucturAl SVM (MANTRA), which prediction relies on a pair of latent variables: h^+ (resp. h^-) provides positive (resp. negative) evidence for a given output y. Secondly, we instantiate MANTRA for two different visual recognition tasks: multi-class classification and ranking. For ranking, we propose efficient solutions to exactly solve the inference and the loss-augmented problems. Finally, extensive experiments highlight the relevance of the proposed method: MANTRA outperforms state-of-the art results on five different datasets.

In this paper, we explore the problem of multiview subspace clustering. We introduce a low-rank tensor constraint to explore the complementary information from multiple views and, accordingly, establish a novel method called Low-rank Tensor constrained Multiview Subspace Clustering (LT-MSC). Our method regards the subspace representation matrices of different views as a tensor, which captures dexterously the high order correlations underlying multiview data. Then the tensor is equipped with a low-rank constraint, which models elegantly the cross information among different views, reduces effectually the redundancy of the learned subspace representations, and improves the accuracy of clustering as well. The inference process of the affinity matrix for clustering is formulated as a tensor nuclear norm minimization problem, constrained with an additional L2,1-norm regularizer and some linear equalities. The minimization problem is convex and thus can be solved efficiently by an Augmented Lagrangian Alternating Direction Minimization (AL-ADM) method. Extensive experimental results on four benchmark datasets show the effectiveness of our proposed LT-MSC method.

Patch-based low-rank models have shown effective in exploiting spatial redundancy of natural images especially for the application of image denoising. However, two-dimensional low-rank model can not fully exploit the spatio-temporal correlation in larger data sets such as multispectral images and 3D MRIs. In this work, we propose a novel low-rank tensor approximation framework with Laplacian Scale Mixture (LSM) modeling for multi-frame image denoising. First, similar 3D patches are grouped to form a tensor of d-order and high-order Singular Value Decomposition (HOSVD) is applied to the grouped tensor. Then the task of multiframe image denoising is formulated as a Maximum A Posterior (MAP) estimation problem with the LSM prior for tensor coefficients. Both unknown sparse coefficients and hidden LSM parameters can be efficiently estimated by the method of alternating optimization. Specifically, we have derived closed-form solutions for both subproblems. Experimental results on spectral and dynamic MRI images show that the proposed algorithm can better preserve the sharpness of important image structures and outperform several existing state-of-the-art multiframe denoising methods (e.g., BM4D and tensor dictionary learning).

Approximating non-linear kernels by finite-dimensional feature maps is a popular approach for speeding up training and evaluation of support vector machines or to encode information into efficient match kernels. We propose a novel method of data independent construction of low dimensional feature maps. The problem is cast as a linear program which jointly considers competing objectives: the quality of the approximation and the dimensionality of the feature map. For both shift-invariant and homogeneous kernels the proposed method achieves a better approximations at the same dimensionality or comparable approximations at lower dimensionality of the feature map compared with state-of-the-art methods.