CVPR 2016 论文列表

Bilateral filters have wide spread use due to their edge-preserving properties. The common use case is to manually choose a parametric filter type, usually a Gaussian filter. In this paper, we will generalize the parametrization and in particular derive a gradient descent algorithm so the filter parameters can be learned from data. This derivation allows to learn high dimensional linear filters that operate in sparsely populated feature spaces. We build on the permutohedral lattice construction for efficient filtering. The ability to learn more general forms of high-dimensional filters can be used in several diverse applications. First, we demonstrate the use in applications where single filter applications are desired for runtime reasons. Further, we show how this algorithm can be used to learn the pairwise potentials in densely connected conditional random fields and apply these to different image segmentation tasks. Finally, we introduce layers of bilateral filters in CNN and propose bilateral neural networks for the use of high-dimensional sparse data. This view provides new ways to encode model structure into network architectures. A diverse set of experiments empirically validates the usage of general forms of filters.

We propose using relaxed deep supervision (RDS) within convolutional neural networks for edge detection. The conventional deep supervision utilizes the general ground-truth to guide intermediate predictions. Instead, we build hierarchical supervisory signals with additional relaxed labels to consider the diversities in deep neural networks. We begin by capturing the relaxed labels from simple detectors (e.g. Canny).Then we merge them with the general ground-truth to generate the RDS. Finally we employ the RDS to supervise the edge network following a coarse-to-fine paradigm. These relaxed labels can be seen as some false positives that are difficult to be classified. We consider these false positives in the supervision, and are able to achieve high performance for better edge detection. We compensate for the lack of training images by capturing coarse edge annotations from a large dataset of image segmentations to pretrain the model. Extensive experiments demonstrate that our approach achieves state-of-the-art performance on the well-known BSDS500 dataset (ODS F-score of .792) and obtains superior cross-dataset generalization results on NYUD dataset.

It is a challenging problem to accurately estimate gazes from low-resolution eye images that do not provide fine and detailed features for eyes. Existing methods attempt to establish the mapping between the visual appearance space to the gaze space. Different from the direct regression approach, the reconstruction-based approach represents appearance and gaze via local linear reconstruction in their own spaces. A common treatment is to use the same local reconstruction in the two spaces, i.e., the reconstruction weights in the appearance space are transferred to the gaze space for gaze reconstruction. However, this questionable treatment is taken for granted but has never been justified, leading to significant errors in gaze estimation. This paper is focused on the study of this fundamental issue. It shows that the distance metric in the appearance space needs to be adjusted, before the same reconstruction can be used. A novel method is proposed to learn the metric, such that the affinity structure of the appearance space under this new metric is as close as possible to the affinity structure of the gaze space under the normal Euclidean metric. Furthermore, the local affinity structure invariance is utilized to further regularize the solution to the reconstruction weights, so as to obtain a more robust and accurate solution. Effectiveness of the proposed method is validated and demonstrated through extensive experiments on different subjects.

We study the problem of online prediction for realtime camera planning, where the goal is to predict smooth trajectories that correctly track and frame objects of interest (e.g., players in a basketball game). The conventional approach for training predictors does not directly consider temporal consistency, and often produces undesirable jitter. Although post-hoc smoothing (e.g., via a Kalman filter) can mitigate this issue to some degree, it is not ideal due to overly stringent modeling assumptions (e.g., Gaussian noise). We propose a recurrent decision tree framework that can directly incorporate temporal consistency into a data-driven predictor, as well as a learning algorithm that can efficiently learn such temporally smooth models. Our approach does not require any post-processing, making online smooth predictions much easier to generate when the noise model is unknown. We apply our approach to sports broadcasting: given noisy player detections, we learn where the camera should look based on human demonstrations. Our experiments exhibit significant improvements over conventional baselines and showcase the practicality of our approach.

We propose a novel visual tracking algorithm based on the representations from a discriminatively trained Convolutional Neural Network (CNN). Our algorithm pretrains a CNN using a large set of videos with tracking ground-truths to obtain a generic target representation. Our network is composed of shared layers and multiple branches of domain-specific layers, where domains correspond to individual training sequences and each branch is responsible for binary classification to identify target in each domain. We train each domain in the network iteratively to obtain generic target representations in the shared layers. When tracking a target in a new sequence, we construct a new network by combining the shared layers in the pretrained CNN with a new binary classification layer, which is updated online. Online tracking is performed by evaluating the candidate windows randomly sampled around the previous target state. The proposed algorithm illustrates outstanding performance in existing tracking benchmarks.

Recent innovations in training deep convolutional neural network (ConvNet) models have motivated the design of new methods to automatically learn local image descriptors. The latest deep ConvNets proposed for this task consist of a siamese network that is trained by penalising misclassification of pairs of local image patches. Current results from machine learning show that replacing this siamese by a triplet network can improve the classification accuracy in several problems, but this has yet to be demonstrated for local image descriptor learning. Moreover, current siamese and triplet networks have been trained with stochastic gradient descent that computes the gradient from individual pairs or triplets of local image patches, which can make them prone to overfitting. In this paper, we first propose the use of triplet networks for the problem of local image descriptor learning. Furthermore, we also propose the use of a global loss that minimises the overall classification error of all patches present in the training set, which can improve the generalisation capability of the model. Using the UBC benchmark dataset for comparing local image descriptors, we show that the triplet network produces a more accurate embedding than the siamese network in terms of the UBC dataset errors. Moreover, we also demonstrate that a combination of the triplet and global losses produces the best embedding in the field, using this triplet network. Finally, we also show that the use of the central-surround siamese network trained with the global loss produces the best result of the field on the UBC dataset.

Can we learn about object classes in images by looking at a collection of relevant 3D models? Or if we want to learn about human (inter-)actions in images, can we benefit from videos or abstract illustrations that show these actions? A common aspect of these settings is the availability of additional or privileged data that can be exploited at training time and that will not be available and not of interest at test time. We seek to generalize the learning with privileged information (LUPI) framework, which requires additional information to be defined per image, to the setting where additional information is a data collection about the task of interest. Our framework minimises the distribution mismatch between errors made in images and in privileged data. The proposed method is tested on four publicly available datasets: Image+ClipArt, Image+3Dobject, and Image+Video. Experimental results reveal that our new LUPI paradigm naturally addresses the cross-dataset learning.

Discriminative deep learning approaches have shown impressive results for problems where human-labeled ground truth is plentiful, but what about tasks where labels are difficult or impossible to obtain? This paper tackles one such problem: establishing dense visual correspondence across different object instances. For this task, although we do not know what the ground-truth is, we know it should be consistent across instances of that category. We exploit this consistency as a supervisory signal to train a convolutional neural network to predict cross-instance correspondences between pairs of images depicting objects of the same category. For each pair of training images we find an appropriate 3D CAD model and render two synthetic views to link in with the pair, establishing a correspondence flow 4-cycle. We use ground-truth synthetic-to-synthetic correspondences, provided by the rendering engine, to train a ConvNet to predict synthetic-to-real, real-to-real and real-to-synthetic correspondences that are cycle-consistent with the ground-truth. At test time, no CAD models are required. We demonstrate that our end-to-end trained ConvNet supervised by cycle-consistency outperforms state-of-the-art pairwise matching methods in correspondence-related tasks.

This paper proposes a method for learning joint embeddings of images and text using a two-branch neural network with multiple layers of linear projections followed by nonlinearities. The network is trained using a large margin objective that combines cross-view ranking constraints with within-view neighborhood structure preservation constraints inspired by metric learning literature. Extensive experiments show that our approach gains significant improvements in accuracy for image-to-text and text-to-image retrieval. Our method achieves new state-of-the-art results on the Flickr30K and MSCOCO image-sentence datasets and shows promise on the new task of phrase localization on the Flickr30K Entities dataset.

State-of-the-art methods for zero-shot visual recognition formulate learning as a joint embedding problem of images and side information. In these formulations the current best complement to visual features are attributes: manually-encoded vectors describing shared characteristics among categories. Despite good performance, attributes have limitations: (1) finer-grained recognition requires commensurately more attributes, and (2) attributes do not provide a natural language interface. We propose to overcome these limitations by training neural language models from scratch; i.e. without pre-training and only consuming words and characters. Our proposed models train end-to-end to align with the fine-grained and category-specific content of images. Natural language provides a flexible and compact way of encoding only the salient visual aspects for distinguishing categories. By training on raw text, our model can do inference on raw text as well, providing humans a familiar mode both for annotation and retrieval. Our model achieves strong performance on zero-shot text-based image retrieval and significantly outperforms the attribute-based state-of-the-art for zero-shot classification on the Caltech-UCSD Birds 200-2011 dataset.

Data in vision domain often exhibit highly-skewed class distribution, i.e., most data belong to a few majority classes, while the minority classes only contain a scarce amount of instances. To mitigate this issue, contemporary classification methods based on deep convolutional neural network (CNN) typically follow classic strategies such as class re-sampling or cost-sensitive training. In this paper, we conduct extensive and systematic experiments to validate the effectiveness of these classic schemes for representation learning on class-imbalanced data. We further demonstrate that more discriminative deep representation can be learned by enforcing a deep network to maintain both inter-cluster and inter-class margins. This tighter constraint effectively reduces the class imbalance inherent in the local data neighborhood. We show that the margins can be easily deployed in standard deep learning framework through quintuplet instance sampling and the associated triple-header hinge loss. The representation learned by our approach, when combined with a simple k-nearest neighbor (kNN) algorithm, shows significant improvements over existing methods on both high- and low-level vision classification tasks that exhibit imbalanced class distribution.

In this work, we revisit the global average pooling layer proposed in [13], and shed light on how it explicitly enables the convolutional neural network (CNN) to have remarkable localization ability despite being trained on image-level labels. While this technique was previously proposed as a means for regularizing training, we find that it actually builds a generic localizable deep representation that exposes the implicit attention of CNNs on image. Despite the apparent simplicity of global average pooling, we are able to achieve 37.1% top-5 error for object localization on ILSVRC 2014 without training on any bounding box annotation. We demonstrate that our network is able to localize the discriminative image regions on a variety of tasks despite not being trained for them.

Learning generic and robust feature representations with data from multiple domains for the same problem is of great value, especially for the problems that have multiple datasets but none of them are large enough to provide abundant data variations. In this work, we present a pipeline for learning deep feature representations from multiple domains with Convolutional Neural Networks (CNNs). When training a CNN with data from all the domains, some neurons learn representations shared across several domains, while some others are effective only for a specific one. Based on this important observation, we propose a Domain Guided Dropout algorithm to improve the feature learning procedure. Experiments show the effectiveness of our pipeline and the proposed algorithm. Our methods on the person re-identification problem outperform state-of-the-art methods on multiple datasets by large margins.

While domain adaptation (DA) aims to associate the learning tasks across data domains, heterogeneous domain adaptation (HDA) particularly deals with learning from cross-domain data which are of different types of features. In other words, for HDA, data from source and target domains are observed in separate feature spaces and thus exhibit distinct distributions. In this paper, we propose a novel learning algorithm of Cross-Domain Landmark Selection (CDLS) for solving the above task. With the goal of deriving a domain-invariant feature subspace for HDA, our CDLS is able to identify representative cross-domain data, including the unlabeled ones in the target domain, for performing adaptation. In addition, the adaptation capabilities of such cross-domain landmarks can be determined accordingly. This is the reason why our CDLS is able to achieve promising HDA performance when comparing to state-of-the-art HDA methods. We conduct classification experiments using data across different features, domains, and modalities. The effectiveness of our proposed method can be successfully verified.

In this paper, we propose a new unsupervised deep learning approach called DeepBit to learn compact binary descriptor for efficient visual object matching. Unlike most existing binary descriptors which were designed with random projections or linear hash functions, we develop a deep neural network to learn binary descriptors in a unsupervised manner. We enforce three criterions on binary codes which are learned at the top layer of our network: 1) minimal loss quantization, 2) evenly distributed codes and 3) uncorrelated bits. Then, we learn the parameters of the networks with a back-propagation technique. Experimental results on three different visual analysis tasks including image matching, image retrieval, and object recognition clearly demonstrate the effectiveness of the proposed approach.

Attributes possess appealing properties and benefit many computer vision problems, such as object recognition, learning with humans in the loop, and image retrieval. Whereas the existing work mainly pursues utilizing attributes for various computer vision problems, we contend that the most basic problem---how to accurately and robustly detect attributes from images---has been left under explored. Especially, the existing work rarely explicitly tackles the need that attribute detectors should generalize well across different categories, including those previously unseen. Noting that this is analogous to the objective of multi-source domain generalization, if we treat each category as a domain, we provide a novel perspective to attribute detection and propose to gear the techniques in multi-source domain generalization for the purpose of learning cross-category generalizable attribute detectors. We validate our understanding and approach with extensive experiments on four challenging datasets and three different problems.

People can recognize scenes across many different modalities beyond natural images. In this paper, we investigate how to learn cross-modal scene representations that transfer across modalities. To study this problem, we introduce a new cross-modal scene dataset. While convolutional neural networks can categorize cross-modal scenes well, they also learn an intermediate representation not aligned across modalities, which is undesirable for cross-modal transfer applications. We present methods to regularize cross-modal convolutional neural networks so that they have a shared representation that is agnostic of the modality. Our experiments suggest that our scene representation can help transfer representations across modalities for retrieval. Moreover, our visualizations suggest that units emerge in the shared representation that tend to activate on consistent concepts independently of the modality.

In this work we improve training of temporal deep models to better learn activity progression for activity detection and early detection. Conventionally, when training a Recurrent Neural Network, specifically a Long Short Term Memory (LSTM) model, the training loss only considers classification error. However, we argue that the detection score of the correct activity category or the detection score margin between the correct and incorrect categories should be monotonically non-decreasing as the model observes more of the activity. We design novel ranking losses that directly penalize the model on violation of such monotonicities, which are used together with classification loss in training of LSTM models. Evaluation on ActivityNet shows significant benefits of the proposed ranking losses in both activity detection and early detection tasks.

When people observe and interact with physical spaces, they are able to associate functionality to regions in the environment. Our goal is to automate functional understanding of large spaces by leveraging activity demonstrations recorded from an ego-centric viewpoint. The method we describe enables functionality estimation in both large scenes where people have behaved, as well as novel scenes where no behaviors are available. Our method learns and predicts "Action Maps", which encode the ability for a user to perform activities at various locations. With the usage of an egocentric camera to observe demonstrations, our method scales with the size of the scene without the need for mounting multiple static surveillance cameras, and is well-suited to the task of observing activities up-close. We demonstrate that by capturing appearance-based attributes of the environment and associating these attributes with activity demonstrations, our mathematical framework allows for the prediction of Action Maps in new environments. Additionally, we take a preliminary look at the breadth of applicability of Action Maps by demonstrating a proof-of-concept application in which they are used in concert with activity detections to perform localization.