CVPR 2019 论文列表

In this paper, we introduce a new idea for unsupervised domain adaptation via a remold of Prototypical Networks, which learn an embedding space and perform classification via a remold of the distances to the prototype of each class. Specifically, we present Transferrable Prototypical Networks (TPN) for adaptation such that the prototypes for each class in source and target domains are close in the embedding space and the score distributions predicted by prototypes separately on source and target data are similar. Technically, TPN initially matches each target example to the nearest prototype in the source domain and assigns an example a "pseudo" label. The prototype of each class could then be computed on source-only, target-only and source-target data, respectively. The optimization of TPN is end-to-end trained by jointly minimizing the distance across the prototypes on three types of data and KL-divergence of score distributions output by each pair of the prototypes. Extensive experiments are conducted on the transfers across MNIST, USPS and SVHN datasets, and superior results are reported when comparing to state-of-the-art approaches. More remarkably, we obtain an accuracy of 80.4% of single model on VisDA 2017 dataset.

We present a new research task and a dataset to understand human social interactions via computational methods, to ultimately endow machines with the ability to encode and decode a broad channel of social signals humans use. This research direction is essential to make a machine that genuinely communicates with humans, which we call Social Artificial Intelligence. We first formulate the "social signal prediction" problem as a way to model the dynamics of social signals exchanged among interacting individuals in a data-driven way. We then present a new 3D motion capture dataset to explore this problem, where the broad spectrum of social signals (3D body, face, and hand motions) are captured in a triadic social interaction scenario. Baseline approaches to predict speaking status, social formation, and body gestures of interacting individuals are presented in the defined social prediction framework.

Monocular depth estimation is a challenging task in scene understanding, with the goal to acquire the geometric properties of 3D space from 2D images. Due to the lack of RGB-depth image pairs, unsupervised learning methods aim at deriving depth information with alternative supervision such as stereo pairs. However, most existing works fail to model the geometric structure of objects, which generally results from considering pixel-level objective functions during training. In this paper, we propose SceneNet to overcome this limitation with the aid of semantic understanding from segmentation. Moreover, our proposed model is able to perform region-aware depth estimation by enforcing semantics consistency between stereo pairs. In our experiments, we qualitatively and quantitatively verify the effectiveness and robustness of our model, which produces favorable results against the state-of-the-art approaches do.

This paper focuses on the temporal aspect for recognizing human activities in videos; an important visual cue that has long been undervalued. We revisit the conventional definition of activity and restrict it to Complex Action: a set of one-actions with a weak temporal pattern that serves a specific purpose. Related works use spatiotemporal 3D convolutions with fixed kernel size, too rigid to capture the varieties in temporal extents of complex actions, and too short for long-range temporal modeling. In contrast, we use multi-scale temporal convolutions, and we reduce the complexity of 3D convolutions. The outcome is Timeception convolution layers, which reasons about minute-long temporal patterns, a factor of 8 longer than best related works. As a result, Timeception achieves impressive accuracy in recognizing the human activities of Charades, Breakfast Actions and MultiTHUMOS. Further, we demonstrate that Timeception learns long-range temporal dependencies and tolerate temporal extents of complex actions.

True video understanding requires making sense of non-lambertian scenes where the color of light arriving at the camera sensor encodes information about not just the last object it collided with, but about multiple mediums -- colored windows, dirty mirrors, smoke or rain. Layered video representations have the potential of accurately modelling realistic scenes but have so far required stringent assumptions on motion, lighting and shape. Here we propose a learning-based approach for multi-layered video representation: we introduce novel uncertainty-capturing 3D convolutional architectures and train them to separate blended videos. We show that these models then generalize to single videos, where they exhibit interesting abilities: color constancy, factoring out shadows and separating reflections. We present quantitative and qualitative results on real world videos.

As deep learning continues to make progress for challenging perception tasks, there is increased interest in combining vision, language, and decision-making. Specifically, the Vision and Language Navigation (VLN) task involves navigating to a goal purely from language instructions and visual information without explicit knowledge of the goal. Recent successful approaches have made in-roads in achieving good success rates for this task but rely on beam search, which thoroughly explores a large number of trajectories and is unrealistic for applications such as robotics. In this paper, inspired by the intuition of viewing the problem as search on a navigation graph, we propose to use a progress monitor developed in prior work as a learnable heuristic for search. We then propose two modules incorporated into an end-to-end architecture: 1) A learned mechanism to perform backtracking, which decides whether to continue moving forward or roll back to a previous state (Regret Module) and 2) A mechanism to help the agent decide which direction to go next by showing directions that are visited and their associated progress estimate (Progress Marker). Combined, the proposed approach significantly outperforms current state-of-the-art methods using greedy action selection, with 5% absolute improvement on the test server in success rates, and more importantly 8% on success rates normalized by the path length.

We present a system for learning full body neural avatars, i.e. deep networks that produce full body renderings of a person for varying body pose and varying camera pose. Our system takes the middle path between the classical graphics pipeline and the recent deep learning approaches that generate images of humans using image-to-image translation. In particular, our system estimates an explicit two-dimensional texture map of the model surface. At the same time, it abstains from explicit shape modeling in 3D. Instead, at test time, the system uses a fully-convolutional network to directly map the configuration of body feature points w.r.t. the camera to the 2D texture coordinates of individual pixels in the image frame. We show that such system is capable of learning to generate realistic renderings while being trained on videos annotated with 3D poses and foreground masks. We also demonstrate that maintaining an explicit texture representation helps our system to achieve better generalization compared to systems that use direct image-to-image translation.

We introduce, TextureNet, a neural network architecture designed to extract features from high-resolution signals associated with 3D surface meshes (e.g., color texture maps). The key idea is to utilize a 4-rotational symmetric(4-RoSy) field to define a domain for convolution on a surface. Though 4-RoSy fields have several properties favor-able for convolution on surfaces (low distortion, few singularities, consistent parameterization, etc.), orientations are ambiguous up to 4-fold rotation at any sample point. So, we introduce a new convolutional operator invariant to the4-RoSy ambiguity and use it in a network to extract features from high-resolution signals on geodesic neighborhoods of a surface. In comparison to alternatives, such as PointNet-based methods which lack a notion of orientation, the coherent structure given by these neighborhoods results in significantly stronger features. As an example application, we demonstrate the benefits of our architecture for 3D semantic segmentation of textured 3D meshes. The results show that our method outperforms all existing methods on the basis of mean IoU by a significant margin in both geometry-only(6.4%) and RGB+Geometry (6.9-8.2%) settings.

In this paper, we propose Text2Scene, a model that generates various forms of compositional scene representations from natural language descriptions. Unlike recent works, our method does NOT use Generative Adversarial Networks (GANs). Text2Scene instead learns to sequentially generate objects and their attributes (location, size, appearance, etc) at every time step by attending to different parts of the input text and the current status of the generated scene. We show that under minor modifications, the proposed framework can handle the generation of different forms of scene representations, including cartoon-like scenes, object layouts corresponding to real images, and synthetic images. Our method is not only competitive when compared with state-of-the-art GAN-based methods using automatic metrics and superior based on human judgments but also has the advantage of producing interpretable results.

Contextual information has been shown to be effective in helping solve various image understanding tasks. Previous works have focused on the extraction of contextual information from an image and use it to infer the properties of some object(s) in the image. In this paper, we consider an inverse problem of how to hallucinate missing contextual information from the properties of a few standalone objects. We refer to it as scene context prediction. This problem is difficult as it requires an extensive knowledge of complex and diverse relationships among different objects in natural scenes. We propose a convolutional neural network, which takes as input the properties (i.e., category, shape, and position) of a few standalone objects to predict an object-level scene layout that compactly encodes the semantics and structure of the scene context where the given objects are. Our quantitative experiments and user studies show that our model can generate more plausible scene context than the baseline approach. We demonstrate that our model allows for the synthesis of realistic scene images from just partial scene layouts and internally learns useful features for scene recognition.

Compared with standard supervised learning, the key difficulty in semi-supervised learning is how to make full use of the unlabeled data. A recently proposed method, virtual adversarial training (VAT), smartly performs adversarial training without label information to impose a local smoothness on the classifier, which is especially beneficial to semi-supervised learning. In this work, we propose tangent-normal adversarial regularization (TNAR) as an extension of VAT by taking the data manifold into consideration. The proposed TNAR is composed by two complementary parts, the tangent adversarial regularization (TAR) and the normal adversarial regularization (NAR). In TAR, VAT is applied along the tangent space of the data manifold, aiming to enforce local invariance of the classifier on the manifold, while in NAR, VAT is performed on the normal space orthogonal to the tangent space, intending to impose robustness on the classifier against the noise causing the observed data deviating from the underlying data manifold. Demonstrated by experiments on both artificial and practical datasets, our proposed TAR and NAR complement with each other, and jointly outperforms other state-of-the-art methods for semi-supervised learning.

In this paper, we provide a modern synthesis of the classic inverse compositional algorithm for dense image alignment. We first discuss the assumptions made by this well-established technique, and subsequently propose to relax these assumptions by incorporating data-driven priors into this model. More specifically, we unroll a robust version of the inverse compositional algorithm and replace multiple components of this algorithm using more expressive models whose parameters we train in an end-to-end fashion from data. Our experiments on several challenging 3D rigid motion estimation tasks demonstrate the advantages of combining optimization with learning-based techniques, outperforming the classic inverse compositional algorithm as well as data-driven image-to-pose regression approaches.

We consider the problem of unsupervised domain adaptation in semantic segmentation. The key in this campaign consists in reducing the domain shift, i.e., enforcing the data distributions of the two domains to be similar. A popular strategy is to align the marginal distribution in the feature space through adversarial learning. However, this global alignment strategy does not consider the local category-level feature distribution. A possible consequence of the global movement is that some categories which are originally well aligned between the source and target may be incorrectly mapped. To address this problem, this paper introduces a category-level adversarial network, aiming to enforce local semantic consistency during the trend of global alignment. Our idea is to take a close look at the category-level data distribution and align each class with an adaptive adversarial loss. Specifically, we reduce the weight of the adversarial loss for category-level aligned features while increasing the adversarial force for those poorly aligned. In this process, we decide how well a feature is category-level aligned between source and target by a co-training approach. In two domain adaptation tasks, i.e., GTA5 -> Cityscapes and SYNTHIA -> Cityscapes, we validate that the proposed method matches the state of the art in segmentation accuracy.

We present the Frontier Aware Search with backTracking (FAST) Navigator, a general framework for action decoding, that achieves state-of-the-art results on the 2018 Room-to-Room (R2R) Vision-and-Language navigation challenge. Given a natural language instruction and photo-realistic image views of a previously unseen environment, the agent was tasked with navigating from source to target location as quickly as possible. While all current approaches make local action decisions or score entire trajectories using beam search, ours balances local and global signals when exploring an unobserved environment. Importantly, this lets us act greedily but use global signals to backtrack when necessary. Applying FAST framework to existing state-of-the-art models achieved a 17% relative gain, an absolute 6% gain on Success rate weighted by Path Length.

Fitting geometric primitives to 3D point cloud data bridges a gap between low-level digitized 3D data and high-level structural information on the underlying 3D shapes. As such, it enables many downstream applications in 3D data processing. For a long time, RANSAC-based methods have been the gold standard for such primitive fitting problems, but they require careful per-input parameter tuning and thus do not scale well for large datasets with diverse shapes. In this work, we introduce Supervised Primitive Fitting Network (SPFN), an end-to-end neural network that can robustly detect a varying number of primitives at different scales without any user control. The network is supervised using ground truth primitive surfaces and primitive membership for the input points. Instead of directly predicting the primitives, our architecture first predicts per-point properties and then uses a differential model estimation module to compute the primitive type and parameters. We evaluate our approach on a novel benchmark of ANSI 3D mechanical component models and demonstrate a significant improvement over both the state-of-the-art RANSAC-based methods and the direct neural prediction.

In this paper, we investigate the issue of knowledge distillation for training compact semantic segmentation networks by making use of cumbersome networks. We start from the straightforward scheme, pixel-wise distillation, which applies the distillation scheme originally introduced for image classification and performs knowledge distillation for each pixel separately. We further propose to distill the structured knowledge from cumbersome networks into compact networks, which is motivated by the fact that semantic segmentation is a structured prediction problem. We study two such structured distillation schemes: (i) pair-wise distillation that distills the pairwise similarities, and (ii) holistic distillation that uses adversarial training to distill holistic knowledge. The effectiveness of our knowledge distillation approaches is demonstrated by extensive experiments on three scene parsing datasets: Cityscapes, Camvid and ADE20K.

Learning unbiased models on imbalanced datasets is a significant challenge. Rare classes tend to get a concentrated representation in the classification space which hampers the generalization of learned boundaries to new test examples. In this paper, we demonstrate that the Bayesian uncertainty estimates directly correlate with the rarity of classes and the difficulty level of individual samples. Subsequently, we present a novel framework for uncertainty based class imbalance learning that follows two key insights: First, classification boundaries should be extended further away from a more uncertain (rare) class to avoid over-fitting and enhance its generalization. Second, each sample should be modeled as a multi-variate Gaussian distribution with a mean vector and a covariance matrix defined by the sample's uncertainty. The learned boundaries should respect not only the individual samples but also their distribution in the feature space. Our proposed approach efficiently utilizes sample and class uncertainty information to learn robust features and more generalizable classifiers. We systematically study the class imbalance problem and derive a novel loss formulation for max-margin learning based on Bayesian uncertainty measure. The proposed method shows significant performance improvements on six benchmark datasets for face verification, attribute prediction, digit/object classification and skin lesion detection.

Dense video captioning is an extremely challenging task since accurate and coherent description of events in a video requires holistic understanding of video contents as well as contextual reasoning of individual events. Most existing approaches handle this problem by first detecting event proposals from a video and then captioning on a subset of the proposals. As a result, the generated sentences are prone to be redundant or inconsistent since they fail to consider temporal dependency between events. To tackle this challenge, we propose a novel dense video captioning framework, which models temporal dependency across events in a video explicitly and leverages visual and linguistic context from prior events for coherent storytelling. This objective is achieved by 1) integrating an event sequence generation network to select a sequence of event proposals adaptively, and 2) feeding the sequence of event proposals to our sequential video captioning network, which is trained by reinforcement learning with two-level rewards---at both event and episode levels---for better context modeling. The proposed technique achieves outstanding performances on ActivityNet Captions dataset in most metrics.

Hair is one of the most challenging objects to reconstruct due to its micro-scale structure and a large number of repeated strands with heavy occlusions. In this paper, we present the first method to capture high-fidelity hair geometry with strand-level accuracy. Our method takes three stages to achieve this. In the first stage, a new multi-view stereo method with a slanted support line is proposed to solve the hair correspondences between different views. In detail, we contribute a novel cost function consisting of both photo-consistency term and geometric term that reconstructs each hair pixel as a 3D line. By merging all the depth maps, a point cloud, as well as local line directions for each point, is obtained. Thus, in the second stage, we feature a novel strand reconstruction method with the mean-shift to convert the noisy point data to a set of strands. Lastly, we grow the hair strands with multi-view geometric constraints to elongate the short strands and recover the missing strands, thus significantly increasing the reconstruction completeness. We evaluate our method on both synthetic data and real captured data, showing that our method can reconstruct hair strands with sub-millimeter accuracy.