CVPR 2017 论文列表

Rotoscoping, the detailed delineation of scene elements through a video shot, is a painstaking task of tremendous importance in professional post-production pipelines. While pixel-wise segmentation techniques can help for this task, professional rotoscoping tools rely on parametric curves that offer the artists a much better interactive control on the definition, editing and manipulation of the segments of interest. Sticking to this prevalent rotoscoping paradigm, we propose a novel framework to capture and track the visual aspect of an arbitrary object in a scene, given a first closed outline of this object. This model combines a collection of local foreground/background appearance models spread along the outline, a global appearance model of the enclosed object and a set of distinctive foreground landmarks. The structure of this rich appearance model allows simple initialization, efficient iterative optimization with exact minimization at each step, and on-line adaptation in videos. We demonstrate qualitatively and quantitatively the merit of this framework through comparisons with tools based on either dynamic segmentation with a closed curve or pixel-wise binary labelling.

The paucity of videos in current action classification datasets (UCF-101 and HMDB-51) has made it difficult to identify good video architectures, as most methods obtain similar performance on existing small-scale benchmarks. This paper re-evaluates state-of-the-art architectures in light of the new Kinetics Human Action Video dataset. Kinetics has two orders of magnitude more data, with 400 human action classes and over 400 clips per class, and is collected from realistic, challenging YouTube videos. We provide an analysis on how current architectures fare on the task of action classification on this dataset and how much performance improves on the smaller benchmark datasets after pre-training on Kinetics. We also introduce a new Two-Stream Inflated 3D ConvNet (I3D) that is based on 2D ConvNet inflation: filters and pooling kernels of very deep image classification ConvNets are expanded into 3D, making it possible to learn seamless spatio-temporal feature extractors from video while leveraging successful ImageNet architecture designs and even their parameters. We show that, after pre-training on Kinetics, I3D models considerably improve upon the state-of-the-art in action classification, reaching 80.2% on HMDB-51 and 97.9% on UCF-101.

Recent years have witnessed a resurgence of interest in video summarization. However, one of the main obstacles to the research on video summarization is the user subjectivity --- users have various preferences over the summaries. The subjectiveness causes at least two problems. First, no single video summarizer fits all users unless it interacts with and adapts to the individual users. Second, it is very challenging to evaluate the performance of a video summarizer. To tackle the first problem, we explore the recently proposed query-focused video summarization which introduces user preferences in the form of text queries about the video into the summarization process. We propose a memory network parameterized sequential determinantal point process in order to attend the user query onto different video frames and shots. To address the second challenge, we contend that a good evaluation metric for video summarization should focus on the semantic information that humans can perceive rather than the visual features or temporal overlaps. To this end, we collect dense per-video-shot concept annotations, compile a new dataset, and suggest an efficient evaluation method defined upon the concept annotations. We conduct extensive experiments contrasting our video summarizer to existing ones and present detailed analyses about the dataset and the new evaluation method.

This paper targets on the problem of set to set recognition, which learns the metric between two image sets. Images in each set belong to the same identity. Since images in a set can be complementary, they hopefully lead to higher accuracy in practical applications. However, the quality of each sample cannot be guaranteed, and samples with poor quality will hurt the metric. In this paper, the quality aware network (QAN) is proposed to confront this problem, where the quality of each sample can be automatically learned although such information is not explicitly provided in the training stage. The network has two branches, where the first branch extracts appearance feature embedding for each sample and the other branch predicts quality score for each sample. Features and quality scores of all samples in a set are then aggregated to generate the final feature embedding. We show that the two branches can be trained in an end-to-end manner given only the set-level identity annotation. Analysis on gradient spread of this mechanism indicates that the quality learned by the network is beneficial to set-to-set recognition and simplifies the distribution that the network needs to fit. Experiments on both face verification and person re-identification show advantages of the proposed QAN. The source code and network structure can be downloaded at GitHub.

Several machine learning tasks require to represent the data using only a sparse set of interest points. An ideal detector is able to find the corresponding interest points even if the data undergo a transformation typical for a given domain. Since the task is of high practical interest in computer vision, many hand-crafted solutions were proposed. In this paper, we ask a fundamental question: can we learn such detectors from scratch? Since it is often unclear what points are "interesting", human labelling cannot be used to find a truly unbiased solution. Therefore, the task requires an unsupervised formulation. We are the first to propose such a formulation: training a neural network to rank points in a transformation-invariant manner. Interest points are then extracted from the top/bottom quantiles of this ranking. We validate our approach on two tasks: standard RGB image interest point detection and challenging cross-modal interest point detection between RGB and depth images. We quantitatively show that our unsupervised method performs better or on-par with baselines.

Scene parsing is challenging for unrestricted open vocabulary and diverse scenes. In this paper, we exploit the capability of global context information by different-region-based context aggregation through our pyramid pooling module together with the proposed pyramid scene parsing network (PSPNet). Our global prior representation is effective to produce good quality results on the scene parsing task, while PSPNet provides a superior framework for pixel-level prediction. The proposed approach achieves state-of-the-art performance on various datasets. It came first in ImageNet scene parsing challenge 2016, PASCAL VOC 2012 benchmark and Cityscapes benchmark. A single PSPNet yields the new record of mIoU accuracy 85.4% on PASCAL VOC 2012 and accuracy 80.2% on Cityscapes.

In this work, we introduce a new kind of spatial partition trees for efficient nearest-neighbor search. Our approach first identifies a set of useful data splitting directions, and then learns a codebook that can be used to encode such directions. We use the product-quantization idea in order to make the effective codebook large, the evaluation of scalar products between the query and the encoded splitting direction very fast, and the encoding itself compact. As a result, the proposed data srtucture (Product Split tree) achieves compact clustering of data points, while keeping the traversal very efficient. In the nearest-neighbor search experiments on high-dimensional data, product split trees achieved state-of-the-art performance, demonstrating better speed-accuracy tradeoff than other spatial partition trees.

Many algorithms for the computation of correspondences between deformable shapes rely on some variant of nearest neighbor matching in a descriptor space. Such are, for example, various point-wise correspondence recovery algorithms used as a post-processing stage in the functional correspondence framework. Such frequently used techniques implicitly make restrictive assumptions (e.g., nearisometry) on the considered shapes and in practice suffer from lack of accuracy and result in poor surjectivity. We propose an alternative recovery technique capable of guaranteeing a bijective correspondence and producing significantly higher accuracy and smoothness. Unlike other methods our approach does not depend on the assumption that the analyzed shapes are isometric. We derive the proposed method from the statistical framework of kernel density estimation and demonstrate its performance on several challenging deformable 3D shape matching datasets.

Deep learning for human action recognition in videos is making significant progress, but is slowed down by its dependency on expensive manual labeling of large video collections. In this work, we investigate the generation of synthetic training data for action recognition, as it has recently shown promising results for a variety of other computer vision tasks. We propose an interpretable parametric generative model of human action videos that relies on procedural generation and other computer graphics techniques of modern game engines. We generate a diverse, realistic, and physically plausible dataset of human action videos, called PHAV for "Procedural Human Action Videos". It contains a total of 39,982 videos, with more than 1,000 examples for each action of 35 categories. Our approach is not limited to existing motion capture sequences, and we procedurally define 14 synthetic actions. We introduce a deep multi-task representation learning architecture to mix synthetic and real videos, even if the action categories differ. Our experiments on the UCF101 and HMDB51 benchmarks suggest that combining our large set of synthetic videos with small real-world datasets can boost recognition performance, significantly outperforming fine-tuning state-of-the-art unsupervised generative models of videos.

A novel algorithm to segment a primary object in a video sequence is proposed in this work. First, we generate candidate regions for the primary object using both color and motion edges. Second,we estimate initial primary object regions, by exploiting the recurrence property of the primary object. Third, we augment the initial regions with missing parts or reducing them by excluding noisy parts repeatedly. This augmentation and reduction process (ARP) identifies the primary object region in each frame. Experimental results demonstrate that the proposed algorithm significantly outperforms the state-of-the-art conventional algorithms on recent benchmark datasets.

While deep feature learning has revolutionized techniques for static-image understanding, the same does not quite hold for video processing. Architectures and optimization techniques used for video are largely based off those for static images, potentially underutilizing rich video information. In this work, we rethink both the underlying network architecture and the stochastic learning paradigm for temporal data. To do so, we draw inspiration from classic theory on linear dynamic systems for modeling time series. By extending such models to include nonlinear mappings, we derive a series of novel recurrent neural networks that sequentially make top-down predictions about the future and then correct those predictions with bottom-up observations. Predictive-corrective networks have a number of desirable properties: (1) they can adaptively focus computation on "surprising" frames where predictions require large corrections, (2) they simplify learning in that only "residual-like" corrective terms need to be learned over time and (3) they naturally decorrelate an input data stream in a hierarchical fashion, producing a more reliable signal for learning at each layer of a network. We provide an extensive analysis of our lightweight and interpretable framework, and demonstrate that our model is competitive with the two-stream network on three challenging datasets without the need for computationally expensive optical flow.

Although the recent success of convolutional neural network (CNN) advances state-of-the-art saliency prediction in static images, few work has addressed the problem of predicting attention in videos. On the other hand, we find that the attention of different subjects consistently focuses on a single face in each frame of videos involving multiple faces. Therefore, we propose in this paper a novel deep learning (DL) based method to predict salient face in multiple-face videos, which is capable of learning features and transition of salient faces across video frames. In particular, we first learn a CNN for each frame to locate salient face. Taking CNN features as input, we develop a multiple-stream long short-term memory (M-LSTM) network to predict the temporal transition of salient faces in video sequences. To evaluate our DL-based method, we build a new eye-tracking database of multiple-face videos. The experimental results show that our method outperforms the prior state-of-the-art methods in predicting visual attention on faces in multiple-face videos.

We introduce a novel strategy for learning to extract semantically meaningful features from aerial imagery. Instead of manually labeling the aerial imagery, we propose to predict (noisy) semantic features automatically extracted from co-located ground imagery. Our network architecture takes an aerial image as input, extracts features using a convolutional neural network, and then applies an adaptive transformation to map these features into the ground-level perspective. We use an end-to-end learning approach to minimize the difference between the semantic segmentation extracted directly from the ground image and the semantic segmentation predicted solely based on the aerial image. We show that a model learned using this strategy, with no additional training, is already capable of rough semantic labeling of aerial imagery. Furthermore, we demonstrate that by finetuning this model we can achieve more accurate semantic segmentation than two baseline initialization strategies. We use our network to address the task of estimating the geolocation and geo-orientation of a ground image. Finally, we show how features extracted from an aerial image can be used to hallucinate a plausible ground-level panorama.

Today, experiencing virtual reality (VR) is a cumbersome experience which either requires dedicated infrastructure like infrared cameras to track the headset and hand-motion controllers (e.g., Oculus Rift, HTC Vive), or provides only 3-DoF (Degrees of Freedom) tracking which severely limits the user experience (e.g., Samsung Gear). To truly enable VR everywhere, we need position tracking to be available as a ubiquitous service. This paper presents WiCapture, a novel approach which leverages commodity WiFi infrastructure, which is ubiquitous today, for tracking purposes. We prototype WiCapture using off-the-shelf WiFi radios and show that it achieves an accuracy of 0.88 cm compared to sophisticated infrared-based tracking systems like the Oculus, while providing much higher range, resistance to occlusion, ubiquity and ease of deployment.

In this work, we introduce the challenging problem of joint multi-person pose estimation and tracking of an unknown number of persons in unconstrained videos. Existing methods for multi-person pose estimation in images cannot be applied directly to this problem, since it also requires to solve the problem of person association over time in addition to the pose estimation for each person. We therefore propose a novel method that jointly models multi-person pose estimation and tracking in a single formulation. To this end, we represent body joint detections in a video by a spatio-temporal graph and solve an integer linear program to partition the graph into sub-graphs that correspond to plausible body pose trajectories for each person. The proposed approach implicitly handles occlusion and truncation of persons. Since the problem has not been addressed quantitatively in the literature, we introduce a challenging "Multi-Person PoseTrack" dataset, and also propose a completely unconstrained evaluation protocol that does not make any assumptions about the scale, size, location or the number of persons. Finally, we evaluate the proposed approach and several baseline methods on our new dataset.

State-of-the-art computer vision algorithms often achieve efficiency by making discrete choices about which hypotheses to explore next. This allows allocation of computational resources to promising candidates, however, such decisions are non-differentiable. As a result, these algorithms are hard to train in an end-to-end fashion. In this work we propose to learn an efficient algorithm for the task of 6D object pose estimation. Our system optimizes the parameters of an existing state-of-the art pose estimation system using reinforcement learning, where the pose estimation system now becomes the stochastic policy, parametrized by a CNN. Additionally, we present an efficient training algorithm that dramatically reduces computation time. We show empirically that our learned pose estimation procedure makes better use of limited resources and improves upon the state-of-the-art on a challenging dataset. Our approach enables differentiable end-to-end training of complex algorithmic pipelines and learns to make optimal use of a given computational budget.

Person recognition methods that use multiple body regions have shown significant improvements over traditional face-based recognition. One of the primary challenges in full-body person recognition is the extreme variation in pose and view point. In this work, (i) we present an approach that tackles pose variations utilizing multiple models that are trained on specific poses, and combined using pose-aware weights during testing. (ii) For learning a person representation, we propose a network that jointly optimizes a single loss over multiple body regions. (iii) Finally, we introduce new benchmarks to evaluate person recognition in diverse scenarios and show significant improvements over previously proposed approaches on all the benchmarks including the photo album setting of PIPA.

We propose a family of quasi-linear discriminants that outperform current large-margin methods in sliding window visual object detection and open set recognition tasks. In these tasks the classification problems are both numerically imbalanced -- positive (object class) training and test windows are much rarer than negative (non-class) ones -- and geometrically asymmetric -- the positive samples typically form compact, visually-coherent groups while negatives are much more diverse, including anything at all that is not a well-centred sample from the target class. It is difficult to cover such negative classes using training samples, and doubly so in `open set' applications where run-time negatives may stem from classes that were not seen at all during training. So there is a need for discriminants whose decision regions focus on tightly circumscribing the positive class, while still taking account of negatives in zones where the two classes overlap. This paper introduces a family of quasi-linear "polyhedral conic" discriminants whose positive regions are distorted L1 balls. The methods have properties and run-time complexities comparable to linear Support Vector Machines (SVMs), and they can be trained from either binary or positive-only samples using constrained quadratic programs related to SVMs. Our experiments show that they significantly outperform both linear SVMs and existing one-class discriminants on a wide range of object detection, open set recognition and conventional closed-set classification tasks.

A number of studies have shown that increasing the depth or width of convolutional networks is a rewarding approach to improve the performance of image recognition. In our study, however, we observed difficulties along both directions. On one hand, the pursuit for very deep networks is met with a diminishing return and increased training difficulty; on the other hand, widening a network would result in a quadratic growth in both computational cost and memory demand. These difficulties motivate us to explore structural diversity in designing deep networks, a new dimension beyond just depth and width. Specifically, we present a new family of modules, namely the PolyInception, which can be flexibly inserted in isolation or in a composition as replacements of different parts of a network. Choosing PolyInception modules with the guidance of architectural efficiency can improve the expressive power while preserving comparable computational cost. The Very Deep PolyNet, designed following this direction, demonstrates substantial improvements over the state-of-the-art on the ILSVRC 2012 benchmark. Compared to Inception-ResNet-v2, it reduces the top-5 validation error on single crops from 4.9% to 4.25%, and that on multi-crops from 3.7% to 3.45%.