CVPR 2016 论文列表

Complex event detection on unconstrained Internet videos has seen much progress in recent years. However, state-of-the-art performance degrades dramatically when the number of positive training exemplars falls short. Since label acquisition is costly, laborious, and time-consuming, there is a real need to consider the much more challenging semantic event search problem, where no example video is given. In this paper, we present a state-of-the-art event search system without any example videos. Relying on the key observation that events (e.g. dog show) are usually compositions of multiple mid-level concepts (e.g. "dog," "theater," and "dog jumping"), we first train a skip-gram model to measure the relevance of each concept with the event of interest. The relevant concept classifiers then cast votes on the test videos but their reliability, due to lack of labeled training videos, has been largely unaddressed. We propose to combine the concept classifiers based on a principled estimate of their accuracy on the unlabeled test videos. A novel warping technique is proposed to improve the performance and an efficient highly-scalable algorithm is provided to quickly solve the resulting optimization. We conduct extensive experiments on the latest TRECVID MEDTest 2014, MEDTest 2013 and CCV datasets, and achieve state-of-the-art performances.

Affine Correspondences (ACs) are more informative than Point Correspondences (PCs) that are used as input in mainstream algorithms for Structure-from-Motion (SfM). Since ACs enable to estimate models from fewer correspondences, its use can dramatically reduce the number of combinations during the iterative step of sample-and-test that exists in most SfM pipelines. However, using ACs instead of PCs as input for SfM passes by fully understanding the relations between ACs and multi-view geometry, as well as by establishing practical, effective AC-based algorithms. This article is a step forward into this direction, by providing a clear account about how ACs constrain the two-view geometry, and by proposing new algorithms for plane segmentation and visual odometry that compare favourably with respect to methods relying in PCs.

We propose an identity-aware multi-object tracker based on the solution path algorithm. Our tracker not only produces identity-coherent trajectories based on cues such as face recognition, but also has the ability to pinpoint potential tracking errors. The tracker is formulated as a quadratic optimization problem with L0 norm constraints, which we propose to solve with the solution path algorithm. The algorithm successively solves the same optimization problem but under different Lp norm constraints, where p gradually decreases from 1 to 0. Inspired by the success of the solution path algorithm in various machine learning tasks, this strategy is expected to converge to a better local minimum than directly minimizing the hardly solvable L0 norm or the roughly approximated L1 norm constraints. Furthermore, the acquired solution path complies with the "decision making process" of the tracker, which provides more insight to locating potential tracking errors. Experiments show that not only is our proposed tracker effective, but also the solution path enables automatic pinpointing of potential tracking failures, which can be readily utilized in an active learning framework to improve identity-aware multi-object tracking.

Vision-based semantic segmentation in urban scenarios is a key functionality for autonomous driving. Recent revolutionary results of deep convolutional neural networks (DCNNs) foreshadow the advent of reliable classifiers to perform such visual tasks. However, DCNNs require learning of many parameters from raw images; thus, having a sufficient amount of diverse images with class annotations is needed. These annotations are obtained via cumbersome, human labour which is particularly challenging for semantic segmentation since pixel-level annotations are required. In this paper, we propose to use a virtual world to automatically generate realistic synthetic images with pixel-level annotations. Then, we address the question of how useful such data can be for semantic segmentation -- in particular, when using a DCNN paradigm. In order to answer this question we have generated a synthetic collection of diverse urban images, named SYNTHIA, with automatically generated class annotations. We use SYNTHIA in combination with publicly available real-world urban images with manually provided annotations. Then, we conduct experiments with DCNNs that show how the inclusion of SYNTHIA in the training stage significantly improves performance on the semantic segmentation task.

To image in high resolution large and occlusion-prone scenes, a camera must move above and around. Degradation of visibility due to geometric occlusions and distances is exacerbated by scattering, when the scene is in a participating medium. Moreover, underwater and in other media, artificial lighting is needed. Overall, data quality depends on the observed surface, medium and the time-varying poses of the camera and light source. This work proposes to optimize camera and light poses as they move, so that the surface is scanned efficiently and the descattered recovery has the highest quality. The work generalizes the next best view concept of robot vision to scattering media and cooperative movable lighting. It also extends descattering to platforms that move optimally. The optimization criterion is information gain, taken from information theory. We exploit the existence of a prior rough 3D model, since underwater such a model is routinely obtained using sonar. We demonstrate this principle in a scaled-down setup.

Deep learning techniques are renowned for supporting effective transfer learning. However, as we demonstrate, the transferred representations support only a few modes of separation and much of its dimensionality is unutilized. In this work we suggest to learn, in the source domain, multiple orthogonal classifiers. We prove that this leads to a reduced rank representation, which however supports more discriminative directions. Interestingly, the softmax probabilities produced by the multiple classifiers are likely to be identical. Extensive experimental results further demonstrate the effectiveness of our method.

Recent face recognition experiments on a major benchmark LFW show stunning performance--a number of algorithms achieve near to perfect score, surpassing human recognition rates. In this paper, we advocate evaluations at the million scale (LFW includes only 13K photos of 5K people). To this end, we have assembled the MegaFace dataset and created the first MegaFace challenge. Our dataset includes One Million photos that capture more than 690K different individuals. The challenge evaluates performance of algorithms with increasing numbers of "distractors" (going from 10 to 1M) in the gallery set. We present both identification and verification performance, evaluate performance with respect to pose and a person's age, and compare as a function of training data size (#photos and #people). We report results of state of the art and baseline algorithms. The MegaFace dataset, baseline code, and evaluation scripts, are all publicly released for further experimentations at http://megaface.cs.washington.edu.

This paper proposes a novel extremely efficient, fully-parallelizable, task-specific algorithm for the computation of global point-wise correspondences in images and videos. Our algorithm, the Global Patch Collider, is based on detecting unique collisions between image points using a collection of learned tree structures that act as conditional hash functions. In contrast to conventional approaches that rely on pairwise distance computation, our algorithm isolates distinctive pixel pairs that hit the same leaf during traversal through multiple learned tree structures. The split functions stored at the intermediate nodes of the trees are trained to ensure that only visually similar patches or their geometric or photometric transformed versions fall into the same leaf node. The matching process involves passing all pixel positions in the images under analysis through the tree structures. We then compute matches by isolating points that uniquely collide with each other ie. fell in the same empty leaf in multiple trees. Our algorithm is linear in the number of pixels but can be made constant time on a parallel computation architecture as the tree traversal for individual image points is decoupled. We demonstrate the efficacy of our method by using it to perform optical flow matching and stereo matching on some challenging benchmarks. Experimental results show that not only is our method extremely computationally efficient, but it is also able to match or outperform state of the art methods that are much more complex.

Visual understanding of complex urban street scenes is an enabling factor for a wide range of applications. Object detection has benefited enormously from large-scale datasets, especially in the context of deep learning. For semantic urban scene understanding, however, no current dataset adequately captures the complexity of real-world urban scenes. To address this, we introduce Cityscapes, a benchmark suite and large-scale dataset to train and test approaches for pixel-level and instance-level semantic labeling. Cityscapes is comprised of a large, diverse set of stereo video sequences recorded in streets from 50 different cities. 5000 of these images have high quality pixel-level annotations; 20000 additional images have coarse annotations to enable methods that leverage large volumes of weakly-labeled data. Crucially, our effort exceeds previous attempts in terms of dataset size, annotation richness, scene variability, and complexity. Our accompanying empirical study provides an in-depth analysis of the dataset characteristics, as well as a performance evaluation of several state-of-the-art approaches based on our benchmark.

This paper studies the Tensor Robust Principal Component (TRPCA) problem which extends the known Robust PCA to the tensor case. Our model is based on a new tensor Singular Value Decomposition (t-SVD) and its induced tensor tubal rank and tensor nuclear norm. Consider that we have a 3-way tensor X in R^n*n*n_3 such that X=L_0+S_0, where L_0 has low tubal rank and S_0 is sparse. Is that possible to recover both components? In this work, we prove that under certain suitable assumptions, we can recover both the low-rank and the sparse components exactly by simply solving a convex program whose objective is a weighted combination of the tensor nuclear norm and the l1-norm, i.e., min L,E s.t. ||L||_*+lambda||E||_1 s.t. X=L+E. where lambda=1/sqrtmax(n_1,n_2)n_3. Interestingly, TRPCA involves RPCA as a special case when n_3=1 and thus it is a simple and elegant tensor extension of RPCA. Also numerical experiments verify our theory and the application for the image denoising demonstrates the effectiveness of our method.

Due to its wide range of applications, matching between two graphs has been extensively studied and remains an active topic. By contrast, it is still under-exploited on how to jointly match multiple graphs, partly due to its intrinsic computational intractability. In this work, we address this challenging problem in a principled way under the rank-1 tensor approximation framework. In particular, we formulate multi-graph matching as a combinational optimization problem with two main ingredients: unary matching over graph vertices and structure matching over graph edges, both of which across multiple graphs. Then we propose an efficient power iteration solution for the resulted NP-hard optimization problem. The proposed algorithm has several advantages: 1) the intrinsic matching consistency across multiple graphs based on the high-order tensor optimization; 2) the free employment of powerful high-order node affinity; 3) the flexible integration between various types of node affinities and edge/hyper-edge affinities. Experiments on diverse and challenging datasets validate the effectiveness of the proposed approach in comparison with state-of-the-arts.

The conventional sparse model relies on data representation in the form of vectors. It represents the vector-valued or vectorized one dimensional (1D) version of an signal as a highly sparse linear combination of basis atoms from a large dictionary. The 1D modeling, though simple, ignores the inherent structure and breaks the local correlation inside multidimensional (MD) signals. It also dramatically increases the demand of memory as well as computational resources especially when dealing with high dimensional signals. In this paper, we propose a new sparse model TenSR based on tensor for MD data representation along with the corresponding MD sparse coding and MD dictionary learning algorithms. The proposed TenSR model is able to well approximate the structure in each mode inherent in MD signals with a series of adaptive separable structure dictionaries via dictionary learning. The proposed MD sparse coding algorithm by proximal method further reduces the computational cost significantly. Experimental results with real world MD signals, i.e. 3D Multi-spectral images, show the proposed TenSR greatly reduces both the computational and memory costs with competitive performance in comparison with the state-of-the-art sparse representation methods. We believe our proposed TenSR model is a promising way to empower the sparse representation especially for large scale high order signals.

This paper presents an approach for reconstruction of 4D temporally coherent models of complex dynamic scenes. No prior knowledge is required of scene structure or camera calibration allowing reconstruction from multiple moving cameras. Sparse-to-dense temporal correspondence is integrated with joint multi-view segmentation and reconstruction to obtain a complete 4D representation of static and dynamic objects. Temporal coherence is exploited to overcome visual ambiguities resulting in improved reconstruction of complex scenes. Robust joint segmentation and reconstruction of dynamic objects is achieved by introducing a geodesic star convexity constraint. Comparative evaluation is performed on a variety of unstructured indoor and outdoor dynamic scenes with hand-held cameras and multiple people. This demonstrates reconstruction of complete temporally coherent 4D scene models with improved non-rigid object segmentation and shape reconstruction.

In view of the advantages of deep networks in producing useful representation, the generated features of different modality data (such as image, audio) can be jointly learned using Multimodal Restricted Boltzmann Machines (MRBM). Recently, audiovisual speech recognition based the MRBM has attracted much attention, and the MRBM shows its effectiveness in learning the joint representation across audiovisual modalities. However, the built networks have weakness in modeling the multimodal sequence which is the natural property of speech signal. In this paper, we will introduce a novel temporal multimodal deep learning architecture, named as Recurrent Temporal Multimodal RBM (RTMRBM), that models multimodal sequences by transforming the sequence of connected MRBMs into a probabilistic series model. Compared with existing multimodal networks, it's simple and efficient in learning temporal joint representation. We evaluate our model on audiovisual speech datasets, two public (AVLetters and AVLetters2) and one self-build. The experimental results demonstrate that our approach can obviously improve the accuracy of recognition compared with standard MRBM and the temporal model based on conditional RBM. In addition, RTMRBM still outperforms non-temporal multimodal deep networks in the presence of the weakness of long-term dependencies.

Dynamic events are often photographed by a number of people from different viewpoints at different times, resulting in an unconstrained set of images. Finding the corresponding moving features in each of the images allows us to extract information about objects of interest in the scene. Computing correspondence of moving features in such a set of images is considerably more challenging than computing correspondence in video due to possible significant differences in viewpoints and inconsistent timing between image captures. The prediction methods used in video for improving robustness and efficiency are not applicable to a set of still images. In this paper we propose a novel method to predict locations of an approximately linear moving feature point, given a small subset of correspondences and the temporal order of image captures. Our method extends the use of epipolar geometry to divide images into valid and invalid regions, termed Temporal Epipolar Regions (TERs). We formally prove that the location of a feature in a new image is restricted to valid TERs. We demonstrate the effectiveness of our method in reducing the search space for correspondence on both synthetic and challenging real world data, and show the improved matching.

We address temporal action localization in untrimmed long videos. This is important because videos in real applications are usually unconstrained and contain multiple action instances plus video content of background scenes or other activities. To address this challenging issue, we exploit the effectiveness of deep networks in temporal action localization via three segment-based 3D ConvNets: (1) a proposal network identifies candidate segments in a long video that may contain actions; (2) a classification network learns one-vs-all action classification model to serve as initialization for the localization network; and (3) a localization network fine-tunes the learned classification network to localize each action instance. We propose a novel loss function for the localization network to explicitly consider temporal overlap and achieve high temporal localization accuracy. In the end, only the proposal network and the localization network are used during prediction. On two large-scale benchmarks, our approach achieves significantly superior performances compared with other state-of-the-art systems: mAP increases from 1.7% to 7.4% on MEXaction2 and increases from 15.0% to 19.0% on THUMOS 2014.

We investigate the feature design and classification architectures in temporal action localization. This application focuses on detecting and labeling actions in untrimmed videos, which brings more challenge than classifying pre-segmented videos. The major difficulty for action localization is the uncertainty of action occurrence and utilization of information from different scales. Two innovations are proposed to address this issue. First, we propose a Pyramid of Score Distribution Feature (PSDF) to capture the motion information at multiple resolutions centered at each detection window. This novel feature mitigates the influence of unknown action position and duration, and shows significant performance gain over previous detection approaches. Second, inter-frame consistency is further explored by incorporating PSDF into the state-of-the-art Recurrent Neural Networks, which gives additional performance gain in detecting actions in temporally untrimmed videos. We tested our action localization framework on the THUMOS'15 and MPII Cooking Activities Dataset, both of which show a large performance improvement over previous attempts.

While current approaches to action recognition on pre-segmented video clips already achieve high accuracies, temporal action detection is still far from comparably good results. Automatically locating and classifying the relevant action segments in videos of varying lengths proves to be a challenging task. We propose a novel method for temporal action detection including statistical length and language modeling to represent temporal and contextual structure. Our approach aims at globally optimizing the joint probability of three components, a length and language model and a discriminative action model, without making intermediate decisions. The problem of finding the most likely action sequence and the corresponding segment boundaries in an exponentially large search space is addressed by dynamic programming. We provide an extensive evaluation of each model component on Thumos 14, a large action detection dataset, and report state-of-the-art results on three datasets.

In this paper we present a deep neural network topology that incorporates a simple to implement transformation-invariant pooling operator (TI-pooling). This operator is able to efficiently handle prior knowledge on nuisance variations in the data, such as rotation or scale changes. Most current methods usually make use of dataset augmentation to address this issue, but this requires larger number of model parameters and more training data, and results in significantly increased training time and larger chance of under- or overfitting. The main reason for these drawbacks is that that the learned model needs to capture adequate features for all the possible transformations of the input. On the other hand, we formulate features in convolutional neural networks to be transformation-invariant. We achieve that using parallel siamese architectures for the considered transformation set and applying the TI-pooling operator on their outputs before the fully-connected layers. We show that this topology internally finds the most optimal "canonical" instance of the input image for training and therefore limits the redundancy in learned features. This more efficient use of training data results in better performance on popular benchmark datasets with smaller number of parameters when comparing to standard convolutional neural networks with dataset augmentation and to other baselines.