CVPR 2017 论文列表

Training convolutional networks (CNN's) that fit on a single GPU with minibatch stochastic gradient descent has become effective in practice. However, there is still no effective method for training large networks that do not fit in the memory of a few GPU cards, or for parallelizing CNN training. In this work we show that a simple hard mixture of experts model can be efficiently trained to good effect on large scale hashtag (multilabel) prediction tasks. Mixture of experts models are not new [??], but in the past, researchers have had to devise sophisticated methods to deal with data fragmentation. We show empirically that modern weakly supervised data sets are large enough to support naive partitioning schemes where each data point is assigned to a single expert. Because the experts are independent, training them in parallel is easy, and evaluation is cheap for the size of the model. Furthermore, we show that we can use a single decoding layer for all the experts, allowing a unified feature embedding space. We demonstrate that it is feasible (and in fact relatively painless) to train far larger models than could be practically trained with standard CNN architectures, and that the extra capacity can be well used on current datasets.

We present an approach that uses a multi-camera system to train fine-grained detectors for keypoints that are prone to occlusion, such as the joints of a hand. We call this procedure multiview bootstrapping: first, an initial keypoint detector is used to produce noisy labels in multiple views of the hand. The noisy detections are then triangulated in 3D using multiview geometry or marked as outliers. Finally, the reprojected triangulations are used as new labeled training data to improve the detector. We repeat this process, generating more labeled data in each iteration. We derive a result analytically relating the minimum number of views to achieve target true and false positive rates for a given detector. The method is used to train a hand keypoint detector for single images. The resulting keypoint detector runs in realtime on RGB images and has accuracy comparable to methods that use depth sensors. The single view detector, triangulated over multiple views, enables 3D markerless hand motion capture with complex object interactions.

In this paper, we propose a novel benchmark for evaluating local image descriptors. We demonstrate that the existing datasets and evaluation protocols do not specify unambiguously all aspects of evaluation, leading to ambiguities and inconsistencies in results reported in the literature. Furthermore, these datasets are nearly saturated due to the recent improvements in local descriptors obtained by learning them from large annotated datasets. Therefore, we introduce a new large dataset suitable for training and testing modern descriptors, together with strictly defined evaluation protocols in several tasks such as matching, retrieval and classification. This allows for more realistic, and thus more reliable comparisons in different application scenarios. We evaluate the performance of several state-of-the-art descriptors and analyse their properties. We show that a simple normalisation of traditional hand-crafted descriptors can boost their performance to the level of deep learning based descriptors within a realistic benchmarks evaluation.

This paper tackles the problem of efficient and effective object instance search in videos. To effectively capture the relevance between a query and video frames and precisely localize the particular object, we leverage the object proposals to improve the quality of object instance search in videos. However, hundreds of object proposals obtained from each frame could result in unaffordable memory and computational cost. To this end, we present a simple yet effective hierarchical object prototype encoding (HOPE) model to accelerate the object instance search without sacrificing accuracy, which exploits both the spatial and temporal self-similarity property existing in object proposals generated from video frames. We design two types of sphere k-means methods, i.e., spatially-constrained sphere k-means and temporally-constrained sphere k-means to learn frame-level object prototypes and dataset-level object prototypes, respectively. In this way, the object instance search problem is cast to the sparse matrix-vector multiplication problem. Thanks to the sparsity of the codes, both the memory and computational cost are significantly reduced. Experimental results on two video datasets demonstrate that our approach significantly improves the performance of video object instance search over other state-of-the-art fast search schemes.

CNNs have made an undeniable impact on computer vision through the ability to learn high-capacity models with large annotated training sets. One of their remarkable properties is the ability to transfer knowledge from a large source dataset to a (typically smaller) target dataset. This is usually accomplished through fine-tuning a fixed-size network on new target data. Indeed, virtually every contemporary visual recognition system makes use of fine-tuning to transfer knowledge from ImageNet. In this work, we analyze what components and parameters change during fine-tuning, and discover that increasing model capacity allows for more natural model adaptation through fine-tuning. By making an analogy to developmental learning, we demonstrate that growing a CNN with additional units, either by widening existing layers or deepening the overall network, significantly outperforms classic fine-tuning approaches. But in order to properly grow a network, we show that newly-added units must be appropriately normalized to allow for a pace of learning that is consistent with existing units. We empirically validate our approach on several benchmark datasets, producing state-of-the-art results.

Spectral Clustering is one of pioneered clustering methods in machine learning and pattern recognition field. It relies on the spectral decomposition criterion to learn a low-dimensonal embedding of data for a basic clustering algorithm such as the k-means. The recent sparse Spectral clustering (SSC) introduces the sparsity for the similarity in low-dimensional space by enforcing a sparsity-induced penalty, resulting a non-convex optimization, and the solution is calculated through a relaxed convex problem via the standard ADMM (Alternative Direction Method of Multipliers), rather than inferring latent representation from eigen-structure. This paper provides a direct solution as solving a new Grassmann optimization problem. By this way calculating latent embedding becomes part of optmization on manifolds and the recently developed manifold optimization methods can be applied. It turns out the learned new features are not only very informative for clustering, but also more intuitive and effective in visualization after dimensionality reduction. We conduct empirical studies on simulated datasets and several real-world benchmark datasets to validate the proposed methods. Experimental results exhibit the effectiveness of this new manifold-based clustering and dimensionality reduction method.

Long Short-Term Memory (LSTM) networks have shown superior performance in 3D human action recognition due to their power in modeling the dynamics and dependencies in sequential data. Since not all joints are informative for action analysis and the irrelevant joints often bring a lot of noise, we need to pay more attention to the informative ones. However, original LSTM does not have strong attention capability. Hence we propose a new class of LSTM network, Global Context-Aware Attention LSTM (GCA-LSTM), for 3D action recognition, which is able to selectively focus on the informative joints in the action sequence with the assistance of global contextual information. In order to achieve a reliable attention representation for the action sequence, we further propose a recurrent attention mechanism for our GCA-LSTM network, in which the attention performance is improved iteratively. Experiments show that our end-to-end network can reliably focus on the most informative joints in each frame of the skeleton sequence. Moreover, our network yields state-of-the-art performance on three challenging datasets for 3D action recognition.

The Gromov-Hausdorff (GH) distance is traditionally used for measuring distances between metric spaces. It was adapted for non-rigid shape comparison and matching of isometric surfaces, and is defined as the minimal distortion of embedding one surface into the other, while the optimal correspondence can be described as the map that minimizes this distortion. Solving such a minimization is a hard combinatorial problem that requires precomputation and storing of all pairwise geodesic distances for the matched surfaces. A popular way for compact representation of functions on surfaces is by projecting them into the leading eigenfunctions of the Laplace-Beltrami Operator (LBO). When truncated, the basis of the LBO is known to be the optimal for representing functions with bounded gradient in a min-max sense. Methods such as Spectral-GMDS exploit this idea to simplify and efficiently approximate a minimization related to the GH distance by operating in the truncated spectral domain, and obtain state of the art results for matching of nearly isometric shapes. However, when considering only a specific set of functions on the surface, such as geodesic distances, an optimized basis could be considered as an even better alternative. Moreover, current simplifications of approximating the GH distance introduce errors due to low rank approximations and relaxations of the permutation matrices. Here, we define the geodesic distance basis, which is optimal for compact approximation of geodesic distances, in terms of Frobenius norm. We use the suggested basis to extract the Geodesic Distance Descriptor (GDD), which encodes the geodesic distances information as a linear combination of the basis functions. We then show how these ideas can be used to efficiently and accurately approximate the metric spaces matching problem with almost no loss of information. We incorporate recent methods for efficient approximation of the proposed basis and descriptor without actually computing and storing all geodesic distances. These observations are used to construct a very simple and efficient procedure for shape correspondence. Experimental results show that the GDD improves both accuracy and efficiency of state of the art shape matching procedures.

This paper proposes a method for generative learning of hierarchical random field models. The resulting model, which we call the hierarchical sparse FRAME (Filters, Random field, And Maximum Entropy) model, is a generalization of the original sparse FRAME model by decomposing it into multiple parts that are allowed to shift their locations, scales and rotations, so that the resulting model becomes a hierarchical deformable template. The model can be trained by an EM-type algorithm that alternates the following two steps: (1) Inference: Given the current model, we match it to each training image by inferring the unknown locations, scales, and rotations of the object and its parts by recursive sum-max maps, and (2) Re-learning: Given the inferred geometric configurations of the objects and their parts, we re-learn the model parameters by maximum likelihood estimation via stochastic gradient algorithm. Experiments show that the proposed method is capable of learning meaningful and interpretable templates that can be used for object detection, classification and clustering.

In this paper, we propose an effective face completion algorithm using a deep generative model. Different from well-studied background completion, the face completion task is more challenging as it often requires to generate semantically new pixels for the missing key components (e.g., eyes and mouths) that contain large appearance variations. Unlike existing nonparametric algorithms that search for patches to synthesize, our algorithm directly generates contents for missing regions based on a neural network. The model is trained with a combination of a reconstruction loss, two adversarial losses and a semantic parsing loss, which ensures pixel faithfulness and local-global contents consistency. With extensive experimental results, we demonstrate qualitatively and quantitatively that our model is able to deal with a large area of missing pixels in arbitrary shapes and generate realistic face completion results.

We present a generative attribute controller (GAC), a novel functionality for generating or editing an image while intuitively controlling large variations of an attribute. This controller is based on a novel generative model called the conditional filtered generative adversarial network (CFGAN), which is an extension of the conventional conditional GAN (CGAN) that incorporates a filtering architecture into the generator input. Unlike the conventional CGAN, which represents an attribute directly using an observable variable (e.g., the binary indicator of attribute presence) so its controllability is restricted to attribute labeling (e.g., restricted to an ON or OFF control), the CFGAN has a filtering architecture that associates an attribute with a multi-dimensional latent variable, enabling latent variations of the attribute to be represented. We also define the filtering architecture and training scheme considering controllability, enabling the variations of the attribute to be intuitively controlled using typical controllers (radio buttons and slide bars). We evaluated our CFGAN on MNIST, CUB, and CelebA datasets and show that it enables large variations of an attribute to be not only represented but also intuitively controlled while retaining identity. We also show that the learned latent space has enough expressive power to conduct attribute transfer and attribute-based image retrieval.

We learn models to generate the immediate future in video. This problem has two main challenges. Firstly, since the future is uncertain, models should be multi-modal, which can be difficult to learn. Secondly, since the future is similar to the past, models store low-level details, which complicates learning of high-level semantics. We propose a framework to tackle both of these challenges. We present a model that generates the future by transforming pixels in the past. Our approach explicitly disentangles the model's memory from the prediction, which helps the model learn desirable invariances. Experiments suggest that this model can generate short videos of plausible futures. We believe predictive models have many applications in robotics, health-care, and video understanding.

Spatial relationships between objects provide important information for text-based image retrieval. As users are more likely to describe a scene from a real world perspective, using 3D spatial relationships rather than 2D relationships that assume a particular viewing direction, one of the main challenges is to infer the 3D structure that bridges images with users' text descriptions. However, direct inference of 3D structure from images requires learning from large scale annotated data. Since interactions between objects can be reduced to a limited set of atomic spatial relations in 3D, we study the possibility of inferring 3D structure from a text description rather than an image, applying physical relation models to synthesize holistic 3D abstract object layouts satisfying the spatial constraints present in a textual description. We present a generic framework for retrieving images from a textual description of a scene by matching images with these generated abstract object layouts. Images are ranked by matching object detection outputs (bounding boxes) to 2D layout candidates (also represented by bounding boxes) which are obtained by projecting the 3D scenes with sampled camera directions. We validate our approach using public indoor scene datasets and show that our method outperforms baselines built upon object occurrence histograms and learned 2D pairwise relations.

Learning how to generate descriptions of images or videos received major interest both in the Computer Vision and Natural Language Processing communities. While a few works have proposed to learn a grounding during the generation process in an unsupervised way (via an attention mechanism), it remains unclear how good the quality of the grounding is and whether it benefits the description quality. In this work we propose a movie description model which learns to generate description and jointly ground (localize) the mentioned characters as well as do visual co-reference resolution between pairs of consecutive sentences/clips. We also propose to use weak localization supervision through character mentions provided in movie descriptions to learn the character grounding. At training time, we first learn how to localize characters by relating their visual appearance to mentions in the descriptions via a semi-supervised approach. We then provide this (noisy) supervision into our description model which greatly improves its performance. Our proposed description model improves over prior work w.r.t. generated description quality and additionally provides grounding and local co-reference resolution. We evaluate it on the MPII Movie Description dataset using automatic and human evaluation measures and using our newly collected grounding and co-reference data for characters.

Due to availability of large amounts of multimedia data, cross-modal matching is gaining increasing importance. Hashing based techniques provide an attractive solution to this problem when the data size is large. Different scenarios of cross-modal matching are possible, for example, data from the different modalities can be associated with a single label or multiple labels, and in addition may or may not have one-to-one correspondence. Most of the existing approaches have been developed for the case where there is one-to-one correspondence between the data of the two modalities. In this paper, we propose a simple, yet effective generalized hashing framework which can work for all the different scenarios, while preserving the semantic distance between the data points. The approach first learns the optimum hash codes for the two modalities simultaneously, so as to preserve the semantic similarity between the data points, and then learns the hash functions to map from the features to the hash codes. Extensive experiments on single label dataset like Wiki and multi-label datasets like NUS-WIDE, Pascal and LabelMe under all the different scenarios and comparisons with the state-of-the-art shows the effectiveness of the proposed approach.

Most popular deep models for action recognition split video sequences into short sub-sequences consisting of a few frames; frame-based features are then pooled for recognizing the activity. Usually, this pooling step discards the temporal order of the frames, which could otherwise be used for better recognition. Towards this end, we propose a novel pooling method, generalized rank pooling (GRP), that takes as input, features from the intermediate layers of a CNN that is trained on tiny sub-sequences, and produces as output the parameters of a subspace which (i) provides a low-rank approximation to the features and (ii) preserves their temporal order. We propose to use these parameters as a compact representation for the video sequence, which is then used in a classification setup. We formulate an objective for computing this subspace as a Riemannian optimization problem on the Grassmann manifold, and propose an efficient conjugate gradient scheme for solving it. Experiments on several activity recognition datasets show that our scheme leads to state-of-the-art performance.

The rational camera model recently introduced in [18] provides a general methodology for studying abstract nonlinear imaging systems and their multi-view geometry. This paper builds on this framework to study "physical realizations" of rational cameras. More precisely, we give an explicit account of the mapping between between physical visual rays and image points (missing in the original description), which allows us to give simple analytical expressions for direct and inverse projections. We also consider "primitive" camera models, that are orbits under the action of various projective transformations, and lead to a general notion of intrinsic parameters. The methodology is general, but it is illustrated concretely by an in-depth study of two-slit cameras, that we model using pairs of linear projections. This simple analytical form allows us to describe models for the corresponding primitive cameras, to introduce intrinsic parameters with a clear geometric meaning, and to define an epipolar tensor characterizing two-view correspondences. In turn, this leads to new algorithms for structure from motion and self-calibration.

Effective integration of local and global contextual information is crucial for dense labeling problems. Most existing methods based on an encoder-decoder architecture simply concatenate features from earlier layers to obtain higher-frequency details in the refinement stages. However, there are limits to the quality of refinement possible if ambiguous information is passed forward. In this paper we propose Gated Feedback Refinement Network (G-FRNet), an end-to-end deep learning framework for dense labeling tasks that addresses this limitation of existing methods. Initially, G-FRNet makes a coarse prediction and then it progressively refines the details by efficiently integrating local and global contextual information during the refinement stages. We introduce gate units that control the information passed forward in order to filter out ambiguity. Experiments on three challenging dense labeling datasets (CamVid, PASCAL VOC 2012, and Horse-Cow Parsing) show the effectiveness of our method. Our proposed approach achieves state-of-the-art results on the CamVid and Horse-Cow Parsing datasets, and produces competitive results on the PASCAL VOC 2012 dataset.

Incorporating smoothness constraints into feature matching is known to enable ultra-robust matching. However, such formulations are both complex and slow, making them unsuitable for video applications. This paper proposes GMS (Grid-based Motion Statistics), a simple means of encapsulating motion smoothness as the statistical likelihood of a certain number of matches in a region. GMS enables translation of high match numbers into high match quality. This provides a real-time, ultra-robust correspondence system. Evaluation on videos, with low textures, blurs and wide-baselines show GMS consistently out-performs other real-time matchers and can achieve parity with more sophisticated, much slower techniques.