ICCV 2017 论文列表

We formulate tracking as an online decision-making process, where a tracking agent must follow an object despite ambiguous image frames and a limited computational budget. Crucially, the agent must decide where to look in the upcoming frames, when to reinitialize because it believes the target has been lost, and when to update its appearance model for the tracked object. Such decisions are typically made heuristically. Instead, we propose to learn an optimal decision-making policy by formulating tracking as a partially observable decision-making process (POMDP). We learn policies with deep reinforcement learning algorithms that need supervision (a reward signal) only when the track has gone awry. We demonstrate that sparse rewards allow us to quickly train on massive datasets, several orders of magnitude more than past work. Interestingly, by treating the data source of Internet videos as unlimited streams, we both learn and evaluate our trackers in a single, unified computational stream.

With an increasing number of users sharing information online, privacy implications entailing such actions are a major concern. For explicit content, such as user profile or GPS data, devices (e.g. mobile phones) as well as web services (e.g. facebook) offer to set privacy settings in order to enforce the users' privacy preferences. We propose the first approach that extends this concept to image content in the spirit of a Visual Privacy Advisor. First, we categorize personal information in images into 68 image attributes and collect a dataset, which allows us to train models that predict such information directly from images. Second, we run a user study to understand the privacy preferences of different users w.r.t. such attributes. Third, we propose models that predict user specific privacy score from images in order to enforce the users' privacy preferences. Our model is trained to predict the user specific privacy risk and even outperforms the judgment of the users, who often fail to follow their own privacy preferences on image data.

Compositional models represent visual patterns as hierarchies of meaningful and reusable parts. They are attractive to vision modeling due to their ability to decompose complex patterns into simpler ones and resolve the low-level ambiguities in high-level image interpretations. However, current compositional models separate structure and part discovery from parameter estimation, which generally leads to suboptimal learning and fitting of the model. Moreover, the commonly adopted latent structural learning is not scalable for deep architectures. To address these difficult issues for compositional models, this paper quests for a unified framework for compositional pattern modeling, inference and learning. Represented by And-Or graphs (AOGs), it jointly models the compositional structure, parts, features, and composition/sub-configuration relationships. We show that the inference algorithm of the proposed framework is equivalent to a feed-forward network. Thus, all the parameters can be learned efficiently via the highly-scalable back-propagation (BP) in an end-to-end fashion. We validate the model via the task of handwritten digit recognition. By visualizing the processes of bottom-up composition and top-down parsing, we show that our model is fully interpretable, being able to learn the hierarchical compositions from visual primitives to visual patterns at increasingly higher levels. We apply this new compositional model to natural scene character recognition and generic object detection. Experimental results have demonstrated its effectiveness.

This paper proposes an accurate and generalizable deep learning framework for iris recognition. The proposed framework is based on a fully convolutional network (FCN), which generates spatially corresponding iris feature descriptors. A specially designed Extended Triplet Loss (ETL) function is introduced to incorporate the bit-shifting and non-iris masking, which are found necessary for learning discriminative spatial iris features. We also developed a sub-network to provide appropriate information for identifying meaningful iris regions, which serves as essential input for the newly developed ETL. Thorough experiments on four publicly available databases suggest that the proposed framework consistently outperforms several classic and state-of-the-art iris recognition approaches. More importantly, our model exhibits superior generalization capability as, unlike popular methods in the literature, it does not essentially require database-specific parameter tuning, which is another key advantage over other approaches.

Despite recent advances in face recognition using deep learning, severe accuracy drops are observed for large pose variations in unconstrained environments. Learning pose-invariant features is one solution, but needs expensively labeled large-scale data and carefully designed feature learning algorithms. In this work, we focus on frontalizing faces in the wild under various head poses, including extreme profile views. We propose a novel deep 3D Morphable Model (3DMM) conditioned Face Frontalization Generative Adversarial Network (GAN), termed as FF-GAN, to generate neutral head pose face images. Our framework differs from both traditional GANs and 3DMM based modeling. Incorporating 3DMM into the GAN structure provides shape and appearance priors for fast convergence with less training data, while also supporting end-to-end training. The 3DMM-conditioned GAN employs not only the discriminator and generator loss but also a new masked symmetry loss to retain visual quality under occlusions, besides an identity loss to recover high frequency information. Experiments on face recognition, landmark localization and 3D reconstruction consistently show the advantage of our frontalization method on faces in the wild datasets.

In this work, we jointly address the problem of text detection and recognition in natural scene images based on convolutional recurrent neural networks. We propose a unified network that simultaneously localizes and recognizes text with a single forward pass, avoiding intermediate processes, such as image cropping, feature re-calculation, word separation, and character grouping. In contrast to existing approaches that consider text detection and recognition as two distinct tasks and tackle them one by one, the proposed framework settles these two tasks concurrently. The whole framework can be trained end-to-end, requiring only images, ground-truth bounding boxes and text labels. The convolutional features are calculated only once and shared by both detection and recognition, which saves processing time. Through multi-task training, the learned features become more informative and improves the overall performance. Our proposed method has achieved competitive performance on several benchmark datasets.

Despite the substantial progress in recent years, the problem of image captioning remains far from being satisfactorily tackled. Sentences produced by existing methods, e.g. those based on LSTM, are often overly rigid and lacking in variability. This issue is related to a learning principle widely used in practice, that is, to maximize the likelihood of training samples. This principle encourages the high resemblance to the "ground-truths", while suppressing other reasonable expressions. Conventional evaluation metrics, e.g. BLEU and METEOR, also favor such restrictive methods. In this paper, we explore an alternative approach, with an aim to improve the naturalness and diversity - two essential properties of human expressions. Specifically, we propose a new framework based on Conditional Generative Adversarial Networks (CGAN), which jointly learns a generator to produce descriptions conditioned on images and an evaluator to assess how well a description fits the visual content. It is noteworthy that training a sequence generator is nontrivial. We overcome the difficulty by Policy Gradient, a strategy stemming from Reinforcement Learning, which allows the generator to receive early feedbacks along the way. We tested our method on two large datasets, where it performed competitively against real people in our user study and outperformed other methods on various tasks.

Recognizing how objects interact with each other is a crucial task in visual recognition. If we define the context of the interaction to be the objects involved, then most current methods can be categorized as either: (i) training a single classifier on the combination of the interaction and its context; or (ii) aiming to recognize the interaction independently of its explicit context. Both methods suffer limitations: the former scales poorly with the number of combinations and fails to generalize to unseen combinations, while the latter often leads to poor interaction recognition performance due to the difficulty of designing a context-independent interaction classifier. To mitigate those drawbacks, this paper proposes an alternative, context-aware interaction recognition framework. The key to our method is to explicitly construct an interaction classifier which combines the context, and the interaction. The context is encoded via word2vec into a semantic space, and is used to derive a classification result for the interaction. The proposed method still builds one classifier for one interaction (as per type (ii) above), but the classifier built is adaptive to context via weights which are context dependent. The benefit of using the semantic space is that it naturally leads to zero-shot generalizations in which semantically similar contexts (subject-object pairs) can be recognized as suitable contexts for an interaction, even if they were not observed in the training set. Our method also scales with the number of interaction-context pairs since our model parameters do not increase with the number of interactions. Thus our method avoids the limitation of both approaches. We demonstrate experimentally that the proposed framework leads to improved performance for all investigated interaction representations and datasets.

In this paper, we study the task of 3D human pose estimation in the wild. This task is challenging due to lack of training data, as existing datasets are either in the wild images with 2D pose or in the lab images with 3D pose. We propose a weakly-supervised transfer learning method that uses mixed 2D and 3D labels in a unified deep neutral network that presents two-stage cascaded structure. Our network augments a state-of-the-art 2D pose estimation sub-network with a 3D depth regression sub-network. Unlike previous two stage approaches that train the two sub-networks sequentially and separately, our training is end-to-end and fully exploits the correlation between the 2D pose and depth estimation sub-tasks. The deep features are better learnt through shared representations. In doing so, the 3D pose labels in controlled lab environments are transferred to in the wild images. In addition, we introduce a 3D geometric constraint to regularize the 3D pose prediction, which is effective in the absence of ground truth depth labels. Our method achieves competitive results on both 2D and 3D benchmarks.

Two types of information exist in a stereo pair: correlation (matching) and decorrelation (half-occlusion). Vision science has shown that both types of information are used in the visual cortex, and that people can perceive depth even when correlation cues are absent or very weak, a capability that remains absent from most computational stereo systems. As a step toward stereo algorithms that are more consistent with these perceptual phenomena, we re-examine the topic of scanline stereo as energy minimization. We represent a disparity profile as a piecewise smooth function with explicit breakpoints between its smooth pieces, and we show this allows correlation and decorrelation to be integrated into an objective that requires only two types of local information: the correlation and its spatial gradient. Experimentally, we show the global optimum of this objective matches human perception on a broad collection of wellknown perceptual stimuli, and that it also provides reasonable piecewise-smooth interpretations of depth in natural images, even without exploiting monocular boundary cues.

In this paper we introduce the TorontoCity benchmark, which covers the full greater Toronto area (GTA) with 712.5km2 of land, 8439km of road and around 400, 000 buildings. Our benchmark provides different perspectives of the world captured from airplanes, drones and cars driving around the city. Manually labeling such a large scale dataset is infeasible. Instead, we propose to utilize different sources of high-precision maps to create our ground truth. Towards this goal, we develop algorithms that allow us to align all data sources with the maps while requiring minimal human supervision. We have designed a wide variety of tasks including building height estimation (reconstruction), road centerline and curb extraction, building instance segmentation, building contour extraction (reorganization), semantic labeling and scene type classification (recognition). Our pilot study shows that most of these tasks are still difficult for modern convolutional neural networks.

We propose an efficient and unified framework, namely ThiNet, to simultaneously accelerate and compress CNN models in both training and inference stages. We focus on the filter level pruning, i.e., the whole filter would be discarded if it is less important. Our method does not change the original network structure, thus it can be perfectly supported by any off-the-shelf deep learning libraries. We formally establish filter pruning as an optimization problem, and reveal that we need to prune filters based on statistics information computed from its next layer, not the current layer, which differentiates ThiNet from existing methods. Experimental results demonstrate the effectiveness of this strategy, which has advanced the state-of-the-art. We also show the performance of ThiNet on ILSVRC-12 benchmark. ThiNet achieves 3.31x FLOPs reduction and 16.63x compression on VGG-16, with only 0.52% top-5 accuracy drop. Similar experiments with ResNet-50 reveal that even for a compact network, ThiNet can also reduce more than half of the parameters and FLOPs, at the cost of roughly 1% top-5 accuracy drop. Moreover, the original VGG-16 model can be further pruned into a very small model with only 5.05MB model size, preserving AlexNet level accuracy but showing much stronger generalization ability.

Current approaches to video forecasting attempt to generate videos directly in pixel space using Generative Adversarial Networks (GANs) or Variational Autoencoders (VAEs). However, since these approaches try to model all the structure and scene dynamics at once, in unconstrained settings they often generate uninterpretable results. Our insight is that forecasting needs to be done first at a higher level of abstraction. Specifically, we exploit human pose detectors as a free source of supervision and break the video forecasting problem into two discrete steps. First we explicitly model the high level structure of active objects in the scene (humans) and use a VAE to model the possible future movements of humans in the pose space. We then use the future poses generated as conditional information to a GAN to predict the future frames of the video in pixel space. By using the structured space of pose as an intermediate representation, we sidestep the problems that GANs have in generating video pixels directly. We show through quantitative and qualitative evaluation that our method outperforms state-of-the-art methods for video prediction.

The Mapillary Vistas Dataset is a novel, large-scale street-level image dataset containing 25,000 high-resolution images annotated into 66 object categories with additional, instance-specific labels for 37 classes. Annotation is performed in a dense and fine-grained style by using polygons for delineating individual objects. Our dataset is 5x larger than the total amount of fine annotations for Cityscapes and contains images from all around the world, captured at various conditions regarding weather, season and daytime. Images come from different imaging devices (mobile phones, tablets, action cameras, professional capturing rigs) and differently experienced photographers. In such a way, our dataset has been designed and compiled to cover diversity, richness of detail and geographic extent. As default benchmark tasks, we define semantic image segmentation and instance-specific image segmentation, aiming to significantly further the development of state-of-the-art methods for visual road-scene understanding.

Neural networks trained on datasets such as ImageNet have led to major advances in visual object classification. One obstacle that prevents networks from reasoning more deeply about complex scenes and situations, and from integrating visual knowledge with natural language, like humans do, is their lack of common sense knowledge about the physical world. Videos, unlike still images, contain a wealth of detailed information about the physical world. However, most labelled video datasets represent high-level concepts rather than detailed physical aspects about actions and scenes. In this work, we describe our ongoing collection of the "something-something" database of video prediction tasks whose solutions require a common sense understanding of the depicted situation. The database currently contains more than 100,000 videos across 174 classes, which are defined as caption-templates. We also describe the challenges in crowd-sourcing this data at scale.

The RPCA model has achieved good performances in various applications. However, two defects limit its effectiveness. Firstly, it is designed for dealing with data in matrix form, which fails to exploit the structure information of higher order tensor data in some pratical situations. Secondly, it adopts L1-norm to tackle noise part which makes it only valid for sparse noise. In this paper, we propose a tensor RPCA model based on CP decomposition and model data noise by Mixture of Gaussians (MoG). The use of tensor structure to raw data allows us to make full use of the inherent structure priors, and MoG is a general approximator to any blends of consecutive distributions, which makes our approach capable of regaining the low dimensional linear subspace from a wide range of noises or their mixture. The model is solved by a new proposed algorithm inferred under a variational Bayesian framework. The superiority of our approach over the existing state-of-the-art approaches is demonstrated by extensive experiments on both of synthetic and real data.

We present a general approach to video understanding, inspired by semantic transfer techniques that have been successfully used for 2D image analysis. Our method considers a video to be a 1D sequence of clips, each one associated with its own semantics. The nature of these semantics -- natural language captions or other labels -- depends on the task at hand. A test video is processed by forming correspondences between its clips and the clips of reference videos with known semantics, following which, reference semantics can be transferred to the test video. We describe two matching methods, both designed to ensure that (a) reference clips appear similar to test clips and (b), taken together, the semantics of the selected reference clips is consistent and maintains temporal coherence. We use our method for video captioning on the LSMDC'16 benchmark, video summarization on the SumMe and TVSum benchmarks, Temporal Action Detection on the Thumos2015 benchmark, and sound prediction on the Greatest Hits benchmark. Our method not only surpasses the state of the art, in four out of five benchmarks, but importantly, it is the only single method we know of that was successfully applied to such a diverse range of tasks.

A temporal superpixel algorithm based on proximity-weighted patch matching (TS-PPM) is proposed in this work. We develop the proximity-weighted patch matching (PPM), which estimates the motion vector of a superpixel robustly, by considering the patch matching distances of neighboring superpixels as well as the target superpixel. In each frame, we initialize superpixels by transferring the superpixel labels of the previous frame using PPM motion vectors. Then, we update the superpixel labels of boundary pixels, based on a cost function, composed of color, spatial, contour, and temporal consistency terms. Finally, we execute superpixel splitting, merging, and relabeling to regularize superpixel sizes and reduce incorrect labels. Experiments show that the proposed algorithm outperforms the state-of-the-art conventional algorithms significantly.

One of the solutions of depth imaging of moving scene is to project a static pattern on the object and use just a single image for reconstruction. However, if the motion of the object is too fast with respect to the exposure time of the image sensor, patterns on the captured image are blurred and reconstruction fails. In this paper, we impose multiple projection patterns into each single captured image to realize temporal super resolution of the depth image sequences. With our method, multiple patterns are projected onto the object with higher fps than possible with a camera. In this case, the observed pattern varies depending on the depth and motion of the object, so we can extract temporal information of the scene from each single image. The decoding process is realized using a learning-based approach where no geometric calibration is needed. Experiments confirm the effectiveness of our method where sequential shapes are reconstructed from a single image. Both quantitative evaluations and comparisons with recent techniques were also conducted.