CVPR 2018 论文列表

Person re-identification (ReID) is to identify pedestrians observed from different camera views based on visual appearance. It is a challenging task due to large pose variations, complex background clutters and severe occlusions. Recently, human pose estimation by predicting joint locations was largely improved in accuracy. It is reasonable to use pose estimation results for handling pose variations and background clutters, and such attempts have obtained great improvement in ReID performance. However, we argue that the pose information was not well utilized and hasn’t yet been fully exploited for person ReID. In this work, we introduce a novel framework called Attention-Aware Compositional Network (AACN) for person ReID. AACN consists of two main components: Pose-guided Part Attention (PPA) and Attention-aware Feature Composition (AFC). PPA is learned and applied to mask out undesirable background features in pedestrian feature maps. Furthermore, pose-guided visibility scores are estimated for body parts to deal with part occlusion in the proposed AFC module. Extensive experiments with ablation analysis show the effectiveness of our method, and state-of-the-art results are achieved on several public datasets, including Market-1501, CUHK03, CUHK01, SenseReID, CUHK03-NP and DukeMTMC-reID.

Recently, substantial research effort has focused on how to apply CNNs or RNNs to better capture temporal patterns in videos, so as to improve the accuracy of video classification. In this paper, however, we show that temporal information, especially longer-term patterns, may not be necessary to achieve competitive results on common trimmed video classification datasets. We investigate the potential of a purely attention based local feature integration. Accounting for the characteristics of such features in video classification, we propose a local feature integration framework based on attention clusters, and introduce a shifting operation to capture more diverse signals. We carefully analyze and compare the effect of different attention mechanisms, cluster sizes, and the use of the shifting operation, and also investigate the combination of attention clusters for multimodal integration. We demonstrate the effectiveness of our framework on three real-world video classification datasets. Our model achieves competitive results across all of these. In particular, on the large-scale Kinetics dataset, our framework obtains an excellent single model accuracy of 79.4% in terms of the top-1 and 94.0% in terms of the top-5 accuracy on the validation set.

Human actions often involve complex interactions across several inter-related objects in the scene. However, existing approaches to fine-grained video understanding or visual relationship detection often rely on single object representation or pairwise object relationships. Furthermore, learning interactions across multiple objects in hundreds of frames for video is computationally infeasible and performance may suffer since a large combinatorial space has to be modeled. In this paper, we propose to efficiently learn higher-order interactions between arbitrary subgroups of objects for fine-grained video understanding. We demonstrate that modeling object interactions significantly improves accuracy for both action recognition and video captioning, while saving more than 3-times the computation over traditional pairwise relationships. The proposed method is validated on two large-scale datasets: Kinetics and ActivityNet Captions. Our SINet and SINet-Caption achieve state-of-the-art performances on both datasets even though the videos are sampled at a maximum of 1 FPS. To the best of our knowledge, this is the first work modeling object interactions on open domain large-scale video datasets, and we additionally model higher-order object interactions which improves the performance with low computational costs.

Vulnerability of Deep Neural Networks (DNNs) to adversarial attacks has been attracting a lot of attention in recent studies. It has been shown that for many state of the art DNNs performing image classification there exist universal adversarial perturbations --- image-agnostic perturbations mere addition of which to natural images with high probability leads to their misclassification. In this work we propose a new algorithm for constructing such universal perturbations. Our approach is based on computing the so-called (p, q)-singular vectors of the Jacobian matrices of hidden layers of a network. Resulting perturbations present interesting visual patterns, and by using only 64 images we were able to construct universal perturbations with more than 60 % fooling rate on the dataset consisting of 50000 images. We also investigate a correlation between the maximal singular value of the Jacobian matrix and the fooling rate of the corresponding singular vector, and show that the constructed perturbations generalize across networks.

The Visual Dialogue task requires an agent to engage in a conversation about an image with a human. It represents an extension of the Visual Question Answering task in that the agent needs to answer a question about an image, but it needs to do so in light of the previous dialogue that has taken place. The key challenge in Visual Dialogue is thus maintaining a consistent, and natural dialogue while continuing to answer questions correctly. We present a novel approach that combines Reinforcement Learning and Generative Adversarial Networks (GANs) to generate more human-like responses to questions. The GAN helps overcome the relative paucity of training data, and the tendency of the typical MLE-based approach to generate overly terse answers. Critically, the GAN is tightly integrated into the attention mechanism that generates human-interpretable reasons for each answer. This means that the discriminative model of the GAN has the task of assessing whether a candidate answer is generated by a human or not, given the provided reason. This is significant because it drives the generative model to produce high quality answers that are well supported by the associated reasoning. The method also generates the state-of-the-art results on the primary benchmark.

This paper introduces a novel method by reshuffling deep features (i.e., permuting the spacial locations of a feature map) of the style image for arbitrary style transfer. We theoretically prove that our new style loss based on reshuffle connects both global and local style losses respectively used by most parametric and non-parametric neural style transfer methods. This simple idea can effectively address the challenging issues in existing style transfer methods. On one hand, it can avoid distortions in local style patterns, and allow semantic-level transfer, compared with neural parametric methods. On the other hand, it can preserve globally similar appearance to the style image, and avoid wash-out artifacts, compared with neural non-parametric methods. Based on the proposed loss, we also present a progressive feature-domain optimization approach. The experiments show that our method is widely applicable to various styles, and produces better quality than existing methods.

Spatiotemporal feature learning in videos is a fundamental problem in computer vision. This paper presents a new architecture, termed as Appearance-and-Relation Network (ARTNet), to learn video representation in an end-to-end manner. ARTNets are constructed by stacking multiple generic building blocks, called as SMART, whose goal is to simultaneously model appearance and relation from RGB input in a separate and explicit manner. Specifically, SMART blocks decouple the spatiotemporal learning module into an appearance branch for spatial modeling and a relation branch for temporal modeling. The appearance branch is implemented based on the linear combination of pixels or filter responses in each frame, while the relation branch is designed based on the multiplicative interactions between pixels or filter responses across multiple frames. We perform experiments on three action recognition benchmarks: Kinetics, UCF101, and HMDB51, demonstrating that SMART blocks obtain an evident improvement over 3D convolutions for spatiotemporal feature learning. Under the same training setting, ARTNets achieve superior performance on these three datasets to the existing state-of-the-art methods.

We present a novel method to train machine learning algorithms to estimate scene depths from a single image, by using the information provided by a camera's aperture as supervision. Prior works use a depth sensor's outputs or images of the same scene from alternate viewpoints as supervision, while our method instead uses images from the same viewpoint taken with a varying camera aperture. To enable learning algorithms to use aperture effects as supervision, we introduce two differentiable aperture rendering functions that use the input image and predicted depths to simulate the depth-of-field effects caused by real camera apertures. We train a monocular depth estimation network end-to-end to predict the scene depths that best explain these finite aperture images as defocus-blurred renderings of the input all-in-focus image.

In this paper, we propose a novel approach for traffic accident anticipation through (i) Adaptive Loss for Early Anticipation (AdaLEA) and (ii) a large-scale self-annotated incident database. The proposed AdaLEA allows us to gradually learn an earlier anticipation as training progresses. The loss function adaptively assigns penalty weights depending on how early the model can anticipate a traffic accident at each epoch. Additionally, a new Near-miss Incident DataBase (NIDB) that contains an enormous number of traffic near-miss incidents in which the four classes of cyclist, pedestrian, vehicle, and background class are labeled is discussed. The NIDB provides joint estimations of traffic incident anticipation and risk-factor categorization. In our experimental results, we found our proposal achieved the highest scores for anticipation (99.1% mean average precision (mAP) and 4.81 sec anticipation of the average time-to-collision (ATTC), values which are +6.6% better and 2.36 sec faster than previous work) and joint estimation (62.1% (mAP) and 3.65 sec anticipation (ATTC), values which are +4.3% better and 0.70 sec faster than previous work).

We consider the problem of segmenting a biomedical image into anatomical regions of interest. We specifically address the frequent scenario where we have no paired training data that contains images and their manual segmentations. Instead, we employ unpaired segmentation images that we use to build an anatomical prior. Critically these segmentations can be derived from imaging data from a different dataset and imaging modality than the current task. We introduce a generative probabilistic model that employs the learned prior through a convolutional neural network to compute segmentations in an unsupervised setting. We conducted an empirical analysis of the proposed approach in the context of structural brain MRI segmentation, using a multi-study dataset of more than 14,000 scans. Our results show that an anatomical prior enables fast unsupervised segmentation which is typically not possible using standard convolutional networks. The integration of anatomical priors can facilitate CNN-based anatomical segmentation in a range of novel clinical problems, where few or no annotations are available and thus standard networks are not trainable. The code, model definitions and model weights are freely available at http://github.com/adalca/neuron

Deep convolutional neural network (ConvNet) is a promising approach for high-performance image classification. The behavior of ConvNet is analyzed mainly based on the neuron activations, such as by visualizing them. In this paper, in contrast to the activations, we focus on filters which are main components of ConvNets. Through analyzing two types of filters at convolution and fully-connected (FC) layers, respectively, on various pre-trained ConvNets, we present the methods to efficiently reformulate the filters, contributing to improving both memory size and classification performance of the ConvNets. They render the filter bases formulated in a parameter-free form as well as the efficient representation for the FC layer. The experimental results on image classification show that the methods are favorably applied to improve various ConvNets, including ResNet, trained on ImageNet with exhibiting high transferability on the other datasets.

Vanishing points and vanishing lines are classical geometrical concepts in perspective cameras that have a lineage dating back to 3 centuries. A vanishing point is a point on the image space where parallel lines in 3D space appear to converge, whereas a vanishing line passes through 2 or more vanishing points. While such concepts are simple and intuitive in perspective cameras, their counterparts in catadioptric cameras (obtained using mirrors and lenses) are more involved. For example, lines in the 3D space map to higher degree curves in catadioptric cameras. The projection of a set of 3D parallel lines converges on a single point in perspective images, whereas they converge to more than one point in catadioptric cameras. To the best of our knowledge, we are not aware of any systematic development of analytical models for vanishing points and vanishing curves in different types of catadioptric cameras. In this paper, we derive parametric equations for vanishing points and vanishing curves using the calibration parameters, mirror shape coefficients, and direction vectors of parallel lines in 3D space. We show compelling experimental results on vanishing point estimation and absolute pose estimation for a wide variety of catadioptric cameras in both simulations and real experiments.

The outstanding performance of deep neural networks (DNNs), for the visual recognition task in particular, has been demonstrated on several large-scale benchmarks. This performance has immensely strengthened the line of re- search that aims to understand and analyze the driving reasons behind the effectiveness of these networks. One important aspect of this analysis has recently gained much attention, namely the reaction of a DNN to noisy input. This has spawned research on developing adversarial input attacks as well as training strategies that make DNNs more robust against these attacks. To this end, we derive in this pa- per exact analytic expressions for the first and second moments (mean and variance) of a small piecewise linear (PL) network (Affine, ReLU, Affine) subject to general Gaussian input. We experimentally show that these expressions are tight under simple linearizations of deeper PL-DNNs, especially popular architectures in the literature (e.g. LeNet and AlexNet). Extensive experiments on image classification show that these expressions can be used to study the behaviour of the output mean of the logits for each class, the interclass confusion and the pixel-level spatial noise sensitivity of the network. Moreover, we show how these expressions can be used to systematically construct targeted and non-targeted adversarial attacks.

A large number of works in egocentric vision have concentrated on action and object recognition. Detection and segmentation of hands in first-person videos, however, has less been explored. For many applications in this domain, it is necessary to accurately segment not only hands of the camera wearer but also the hands of others with whom he is interacting. Here, we take an in-depth look at the hand segmentation problem. In the quest for robust hand segmentation methods, we evaluated the performance of the state of the art semantic segmentation methods, off the shelf and fine-tuned, on existing datasets. We fine-tune RefineNet, a leading semantic segmentation method, for hand segmentation and find that it does much better than the best contenders. Existing hand segmentation datasets are collected in the laboratory settings. To overcome this limitation, we contribute by collecting two new datasets: a) EgoYouTubeHands including egocentric videos containing hands in the wild, and b) HandOverFace to analyze the performance of our models in presence of similar appearance occlusions. We further explore whether conditional random fields can help refine generated hand segmentations. To demonstrate the benefit of accurate hand maps, we train a CNN for hand-based activity recognition and achieve higher accuracy when a CNN was trained using hand maps produced by the fine-tuned RefineNet. Finally, we annotate a subset of the EgoHands dataset for fine-grained action recognition and show that an accuracy of 58.6% can be achieved by just looking at a single hand pose which is much better than the chance level (12.5%).

We present a fast learning-based algorithm for deformable, pairwise 3D medical image registration. Current registration methods optimize an objective function independently for each pair of images, which can be time-consuming for large data. We define registration as a parametric function, and optimize its parameters given a set of images from a collection of interest. Given a new pair of scans, we can quickly compute a registration field by directly evaluating the function using the learned parameters. We model this function using a CNN, and use a spatial transform layer to reconstruct one image from another while imposing smoothness constraints on the registration field. The proposed method does not require supervised information such as ground truth registration fields or anatomical landmarks. We demonstrate registration accuracy comparable to state-of-the-art 3D image registration, while operating orders of magnitude faster in practice. Our method promises to significantly speed up medical image analysis and processing pipelines, while facilitating novel directions in learning-based registration and its applications. Our code is available at https://github.com/balakg/voxelmorph.

Text detection and recognition in natural images have long been considered as two separate tasks that are processed sequentially. Jointly training two tasks is non-trivial due to significant differences in learning difficulties and convergence rates. In this work, we present a conceptually simple yet efficient framework that simultaneously processes the two tasks in a united framework. Our main contributions are three-fold: (1) we propose a novel textalignment layer that allows it to precisely compute convolutional features of a text instance in arbitrary orientation, which is the key to boost the performance; (2) a character attention mechanism is introduced by using character spatial information as explicit supervision, leading to large improvements in recognition; (3) two technologies, together with a new RNN branch for word recognition, are integrated seamlessly into a single model which is end-to-end trainable. This allows the two tasks to work collaboratively by sharing convolutional features, which is critical to identify challenging text instances. Our model obtains impressive results in end-to-end recognition on the ICDAR 2015, significantly advancing the most recent results, with improvements of F-measure from (0.54, 0.51, 0.47) to (0.82, 0.77, 0.63), by using a strong, weak and generic lexicon respectively. Thanks to joint training, our method can also serve as a good detector by achieving a new state-of-the-art detection performance on related benchmarks. Code is available at https://github. com/tonghe90/textspotter.

In this paper, we study the mixture proportion estimation (MPE) problem in a new setting: given samples from the mixture and the component distributions, we identify the proportions of the components in the mixture distribution. To address this problem, we make use of a linear independence assumption, i.e., the component distributions are independent from each other, which is much weaker than assumptions exploited in the previous MPE methods. Based on this assumption, we propose a method (1) that uniquely identifies the mixture proportions, (2) whose output provably converges to the optimal solution, and (3) that is computationally efficient. We show the superiority of the proposed method over the state-of-the-art methods in two applications including learning with label noise and semi-supervised learning on both synthetic and real-world datasets.

An analysis of different techniques for recognizing and detecting objects under extreme scale variation is presented. Scale specific and scale invariant design of detectors are compared by training them with different configurations of input data. By evaluating the performance of different network architectures for classifying small objects on ImageNet, we show that CNNs are not robust to changes in scale. Based on this analysis, we propose to train and test detectors on the same scales of an image-pyramid. Since small and large objects are difficult to recognize at smaller and larger scales respectively, we present a novel training scheme called Scale Normalization for Image Pyramids (SNIP) which selectively back-propagates the gradients of object instances of different sizes as a function of the image scale. On the COCO dataset, our single model performance is 45.7% and an ensemble of 3 networks obtains an mAP of 48.3%. We use off-the-shelf ImageNet-1000 pre-trained models and only train with bounding box supervision. Our submission won the Best Student Entry in the COCO 2017 challenge. Code will be made available at url{http://bit.ly/2yXVg4c}.

One approach to improve the accuracy and robustness of vision-aided inertial navigation systems (VINS) that employ low-cost inertial sensors, is to obtain scale information from stereoscopic vision. Processing images from two cameras, however, is computationally expensive and increases latency. To address this limitation, in this work, a novel two-camera alternating-stereo VINS is presented. Specifically, the proposed system triggers the left-right cameras in an alternating fashion, estimates the poses corresponding to the left camera only, and introduces a linear interpolation model for processing the alternating right camera measurements. Although not a regular stereo system, the alternating visual observations when employing the proposed interpolation scheme, still provide scale information, as shown by analyzing the observability properties of the vision-only corresponding system. Finally, the performance gain, of the proposed algorithm over its monocular and stereo counterparts is assessed using various datasets.