CVPR 2019 Papers

Volumetric (4D) performance capture is fundamental for AR/VR content generation. Whereas previous work in 4D performance capture has shown impressive results in studio settings, the technology is still far from being accessible to a typical consumer who, at best, might own a single RGBD sensor. Thus, in this work, we propose a method to synthesize free viewpoint renderings using a single RGBD camera. The key insight is to leverage previously seen "calibration" images of a given user to extrapolate what should be rendered in a novel viewpoint from the data available in the sensor. Given these past observations from multiple viewpoints, and the current RGBD image from a fixed view, we propose an end-to-end framework that fuses both these data sources to generate novel renderings of the performer. We demonstrate that the method can produce high fidelity images, and handle extreme changes in subject pose and camera viewpoints. We also show that the system generalizes to performers not seen in the training data. We run exhaustive experiments demonstrating the effectiveness of the proposed semi-parametric model (i.e. calibration images available to the neural network) compared to other state of the art machine learned solutions. Further, we compare the method with more traditional pipelines that employ multi-view capture. We show that our framework is able to achieve compelling results, with substantially less infrastructure than previously required.

We introduce the first visual privacy dataset originating from people who are blind in order to better understand their privacy disclosures and to encourage the development of algorithms that can assist in preventing their unintended disclosures. It includes 8,862 regions showing private content across 5,537 images taken by blind people. Of these, 1,403 are paired with questions and 62% of those directly ask about the private content. Experiments demonstrate the utility of this data for predicting whether an image shows private information and whether a question asks about the private content in an image. The dataset is publicly-shared at http://vizwiz.org/data/.

Visual question answering (VQA) demands simultaneous comprehension of both the image visual content and natural language questions. In some cases, the reasoning needs the help of common sense or general knowledge which usually appear in the form of text. Current methods jointly embed both the visual information and the textual feature into the same space. Nevertheless, how to model the complex interactions between the two different modalities is not an easy work. In contrast to struggling on multimodal feature fusion, in this paper, we propose to unify all the input information by natural language so as to convert VQA into a machine reading comprehension problem. With this transformation, our method not only can tackle VQA datasets that focus on observation based questions, but can also be naturally extended to handle knowledge-based VQA which requires to explore large-scale external knowledge base. It is a step towards being able to exploit large volumes of text and natural language processing techniques to address VQA problem. Two types of models are proposed to deal with open-ended VQA and multiple-choice VQA respectively. We evaluate our models on three VQA benchmarks. The comparable performance with the state-of-the-art demonstrates the effectiveness of the proposed method.

Visual Query Answering (VQA) is of great significance in offering people convenience: one can raise a question for details of objects, or high-level understanding about the scene, over an image. This paper proposes a novel method to address the VQA problem. In contrast to prior works, our method that targets single scene VQA, replies on graph-based techniques and involves reasoning. In a nutshell, our approach is centered on three graphs. The first graph, referred to as inference graph G_I, is constructed via learning over labeled data. The other two graphs, referred to as query graph Q and entity-attribute graph EAG, are generated from natural language query NLQ and image Img, that are issued from users, respectively. As EAG often does not take sufficient information to answer Q, we develop techniques to infer missing information of EAG with G_I. Based on EAG and Q, we provide techniques to find matches of Q in EAG, as the answer of NLQ in Img. Unlike commonly used VQA methods that are based on end-to-end neural networks, our graph-based method shows well-designed reasoning capability, and thus is highly interpretable. We also create a dataset on soccer match (Soccer-VQA) with rich annotations. The experimental results show that our approach outperforms the state-of-the-art method and has high potential for future investigation.

We introduce a novel CNN-based approach for visual localization from a single RGB image that relies on densely matching a set of Objects-of-Interest (OOIs). In this paper, we focus on planar objects which are highly descriptive in an environment, such as paintings in museums or logos and storefronts in malls or airports. For each OOI, we define a reference image for which 3D world coordinates are available. Given a query image, our CNN model detects the OOIs, segments them and finds a dense set of 2D-2D matches between each detected OOI and its corresponding reference image. Given these 2D-2D matches, together with the 3D world coordinates of each reference image, we obtain a set of 2D-3D matches from which solving a Perspective-n-Point problem gives a pose estimate. We show that 2D-3D matches for reference images, as well as OOI annotations can be obtained for all training images from a single instance annotation per OOI by leveraging Structure-from-Motion reconstruction. We introduce a novel synthetic dataset, VirtualGallery, which targets challenges such as varying lighting conditions and different occlusion levels. Our results show that our method achieves high precision and is robust to these challenges. We also experiment using the Baidu localization dataset captured in a shopping mall. Our approach is the first deep regression-based method to scale to such a larger environment.

Human visual perception shows good consistency for many multi-label image classification tasks under certain spatial transforms, such as scaling, rotation, flipping and translation. This has motivated the data augmentation strategy widely used in CNN classifier training -- transformed images are included for training by assuming the same class labels as their original images. In this paper, we further propose the assumption of perceptual consistency of visual attention regions for classification under such transforms, i.e., the attention region for a classification follows the same transform if the input image is spatially transformed. While the attention regions of CNN classifiers can be derived as an attention heatmap in middle layers of the network, we find that their consistency under many transforms are not preserved. To address this problem, we propose a two-branch network with an original image and its transformed image as inputs and introduce a new attention consistency loss that measures the attention heatmap consistency between two branches. This new loss is then combined with multi-label image classification loss for network training. Experiments on three datasets verify the superiority of the proposed network by achieving new state-of-the-art classification performance.

We present Vision-based Navigation with Language-based Assistance (VNLA), a grounded vision-language task where an agent with visual perception is guided via language to find objects in photorealistic indoor environments. The task emulates a real-world scenario in that (a) the requester may not know how to navigate to the target objects and thus makes requests by only specifying high-level end-goals, and (b) the agent is capable of sensing when it is lost and querying an advisor, who is more qualified at the task, to obtain language subgoals to make progress. To model language-based assistance, we develop a general framework termed Imitation Learning with Indirect Intervention (I3L), and propose a solution that is effective on the VNLA task. Empirical results show that this approach significantly improves the success rate of the learning agent over other baselines on both seen and unseen environments. Our code and data are publicly available at https://github.com/debadeepta/vnla .

Recent years have witnessed the growing interest in visual quality assessment (VQA) for 360deg video. Unfortunately, the existing VQA approaches do not consider the facts that: 1) Observers only see viewports of 360deg video, rather than patches or whole 360deg frames. 2) Within the viewport, only salient regions can be perceived by observers with high resolution. Thus, this paper proposes a viewport-based convolutional neural network (V-CNN) approach for VQA on 360deg video, considering both auxiliary tasks of viewport proposal and viewport saliency prediction. Our V-CNN approach is composed of two stages, i.e., viewport proposal and VQA. In the first stage, the viewport proposal network (VP-net) is developed to yield several potential viewports, seen as the first auxiliary task. In the second stage, a viewport quality network (VQ-net) is designed to rate the VQA score for each proposed viewport, in which the saliency map of the viewport is predicted and then utilized in VQA score rating. Consequently, another auxiliary task of viewport saliency prediction can be achieved. More importantly, the main task of VQA on 360deg video can be accomplished via integrating the VQA scores of all viewports. The experiments validate the effectiveness of our V-CNN approach in significantly advancing the state-of-the-art performance of VQA on 360deg video. In addition, our approach achieves comparable performance in two auxiliary tasks. The code of our V-CNN approach is available at https://github.com/Archer-Tatsu/V-CNN.

We consider the problem of video summarization. Given an input raw video, the goal is to select a small subset of key frames from the input video to create a shorter summary video that best describes the content of the original video. Most of the current state-of-the-art video summarization approaches use supervised learning and require labeled training data. Each training instance consists of a raw input video and its ground truth summary video curated by human annotators. However, it is very expensive and difficult to create such labeled training examples. To address this limitation, we propose a novel formulation to learn video summarization from unpaired data. We present an approach that learns to generate optimal video summaries using a set of raw videos (V) and a set of summary videos (S), where there exists no correspondence between V and S. We argue that this type of data is much easier to collect. Our model aims to learn a mapping function F : V -> S such that the distribution of resultant summary videos from F(V) is similar to the distribution of S with the help of an adversarial objective. In addition, we enforce a diversity constraint on F(V) to ensure that the generated video summaries are visually diverse. Experimental results on two benchmark datasets indicate that our proposed approach significantly outperforms other alternative methods.

Visual relationship reasoning is a crucial yet challenging task for understanding rich interactions across visual concepts. For example, a relationship \ man, open, door\ involves a complex relation \ open\ between concrete entities \ man, door\ . While much of the existing work has studied this problem in the context of still images, understanding visual relationships in videos has received limited attention. Due to their temporal nature, videos enable us to model and reason about a more comprehensive set of visual relationships, such as those requiring multiple (temporal) observations (e.g., \ man, lift up, box\ vs. \ man, put down, box\ ), as well as relationships that are often correlated through time (e.g., \ woman, pay, money\ followed by \ woman, buy, coffee\ ). In this paper, we construct a Conditional Random Field on a fully-connected spatio-temporal graph that exploits the statistical dependency between relational entities spatially and temporally. We introduce a novel gated energy function parametrization that learns adaptive relations conditioned on visual observations. Our model optimization is computationally efficient, and its space computation complexity is significantly amortized through our proposed parameterization. Experimental results on benchmark video datasets (ImageNet Video and Charades) demonstrate state-of-the-art performance across three standard relationship reasoning tasks: Detection, Tagging, and Recognition.

Video magnification methods can magnify and reveal subtle changes invisible to the naked eye. However, in such subtle changes, meaningful ones caused by physical and natural phenomena are mixed with non-meaningful ones caused by photographic noise. Therefore, current methods often produce noisy and misleading magnification outputs due to the non-meaningful subtle changes. For detecting only meaningful subtle changes, several methods have been proposed but require human manipulations, additional resources, or input video scene limitations. In this paper, we present a novel method using fractional anisotropy (FA) to detect only meaningful subtle changes without the aforementioned requirements. FA has been used in neuroscience to evaluate anisotropic diffusion of water molecules in the body. On the basis of our observation that temporal distribution of meaningful subtle changes more clearly indicates anisotropic diffusion than that of non-meaningful ones, we used FA to design a fractional anisotropic filter that passes only meaningful subtle changes. Using the filter enables our method to obtain better and more impressive magnification results than those obtained with state-of-the-art methods.

This paper proposes the novel task of video generation conditioned on a SINGLE semantic label map, which provides a good balance between flexibility and quality in the generation process. Different from typical end-to-end approaches, which model both scene content and dynamics in a single step, we propose to decompose this difficult task into two sub-problems. As current image generation methods do better than video generation in terms of detail, we synthesize high quality content by only generating the first frame. Then we animate the scene based on its semantic meaning to obtain temporally coherent video, giving us excellent results overall. We employ a cVAE for predicting optical flow as a beneficial intermediate step to generate a video sequence conditioned on the initial single frame. A semantic label map is integrated into the flow prediction module to achieve major improvements in the image-to-video generation process. Extensive experiments on the Cityscapes dataset show that our method outperforms all competing methods.

Most existing ensemble methods aim to train the underlying embedded models independently and simply aggregate their final outputs via averaging or weighted voting. As many prediction tasks contain uncertainty, most of these ensemble methods just reduce variance of the predictions without considering the collaborations among the ensembles. Different from these ensemble methods, multiple choice learning (MCL) methods exploit the cooperation among all the embedded models to generate multiple diverse hypotheses. In this paper, a new MCL method, called vMCL (the abbreviation of versatile Multiple Choice Learning), is developed to extend the application scenarios of MCL methods by ensembling deep neural networks. Our vMCL method keeps the advantage of existing MCL methods while overcoming their major drawback, thus achieves better performance. The novelty of our vMCL lies in three aspects: (1) a choice network is designed to learn the confidence level of each specialist which can provide the best prediction base on multiple hypotheses; (2) a hinge loss is introduced to alleviate the overconfidence issue in MCL settings; (3) Easy to be implemented and can be trained in an end-to-end manner, which is a very attractive feature for many real-world applications. Experiments on image classification and image segmentation task show that vMCL outperforms the existing state-of-the-art MCL methods.

Robust geometric and semantic scene understanding is ever more important in many real-world applications such as autonomous driving and robotic navigation. In this paper, we propose a multi-task learning-based approach capable of jointly performing geometric and semantic scene understanding, namely depth prediction (monocular depth estimation and depth completion) and semantic scene segmentation. Within a single temporally constrained recurrent network, our approach uniquely takes advantage of a complex series of skip connections, adversarial training and the temporal constraint of sequential frame recurrence to produce consistent depth and semantic class labels simultaneously. Extensive experimental evaluation demonstrates the efficacy of our approach compared to other contemporary state-of-the-art techniques.

In daily life, graphic symbols, such as traffic signs and brand logos, are ubiquitously utilized around us due to its intuitive expression beyond language boundary. We tackle an open-set graphic symbol recognition problem by one-shot classification with prototypical images as a single training example for each novel class. We take an approach to learn a generalizable embedding space for novel tasks. We propose a new approach called variational prototyping-encoder (VPE) that learns the image translation task from real-world input images to their corresponding prototypical images as a meta-task. As a result, VPE learns image similarity as well as prototypical concepts which differs from widely used metric learning based approaches. Our experiments with diverse datasets demonstrate that the proposed VPE performs favorably against competing metric learning based one-shot methods. Also, our qualitative analyses show that our meta-task induces an effective embedding space suitable for unseen data representation.

Transferring knowledge from a teacher neural network pretrained on the same or a similar task to a student neural network can significantly improve the performance of the student neural network. Existing knowledge transfer approaches match the activations or the corresponding hand-crafted features of the teacher and the student networks. We propose an information-theoretic framework for knowledge transfer which formulates knowledge transfer as maximizing the mutual information between the teacher and the student networks. We compare our method with existing knowledge transfer methods on both knowledge distillation and transfer learning tasks and show that our method consistently outperforms existing methods. We further demonstrate the strength of our method on knowledge transfer across heterogeneous network architectures by transferring knowledge from a convolutional neural network (CNN) to a multi-layer perceptron (MLP) on CIFAR-10. The resulting MLP significantly outperforms the-state-of-the-art methods and it achieves similar performance to the CNN with a single convolutional layer.

We propose a variational Bayesian scheme for pruning convolutional neural networks in channel level. This idea is motivated by the fact that deterministic value based pruning methods are inherently improper and unstable. In a nutshell, variational technique is introduced to estimate distribution of a newly proposed parameter, called channel saliency, based on this, redundant channels can be removed from model via a simple criterion. The advantages are two-fold: 1) Our method conducts channel pruning without desire of re-training stage, thus improving the computation efficiency. 2) Our method is implemented as a stand-alone module, called variational pruning layer, which can be straightforwardly inserted into off-the-shelf deep learning packages, without any special network design. Extensive experimental results well demonstrate the effectiveness of our method: For CIFAR-10, we perform channel removal on different CNN models up to 74% reduction, which results in significant size reduction and computation saving. For ImageNet, about 40% channels of ResNet-50 are removed without compromising accuracy.

Variational dropout (VD) is a generalization of Gaussian dropout, which aims at inferring the posterior of network weights based on a log-uniform prior on them to learn these weights as well as dropout rate simultaneously. The log-uniform prior not only interprets the regularization capacity of Gaussian dropout in network training, but also underpins the inference of such posterior. However, the log-uniform prior is an improper prior (i.e., its integral is infinite), which causes the inference of posterior to be ill-posed, thus restricting the regularization performance of VD. To address this problem, we present a new generalization of Gaussian dropout, termed variational Bayesian dropout (VBD), which turns to exploit a hierarchical prior on the network weights and infer a new joint posterior. Specifically, we implement the hierarchical prior as a zero-mean Gaussian distribution with variance sampled from a uniform hyper-prior. Then, we incorporate such a prior into inferring the joint posterior over network weights and the variance in the hierarchical prior, with which both the network training and dropout rate estimation can be cast into a joint optimization problem. More importantly, the hierarchical prior is a proper prior which enables the inference of posterior to be well-posed. In addition, we further show that the proposed VBD can be seamlessly applied to network compression. Experiments on classification and network compression demonstrate the superior performance of the proposed VBD in regularizing network training.

The Variational Autoencoder (VAE) is a powerful architecture capable of representation learning and generative modeling. When it comes to learning interpretable (disentangled) representations, VAE and its variants show unparalleled performance. However, the reasons for this are unclear, since a very particular alignment of the latent embedding is needed but the design of the VAE does not encourage it in any explicit way. We address this matter and offer the following explanation: the diagonal approximation in the encoder together with the inherent stochasticity force local orthogonality of the decoder. The local behavior of promoting both reconstruction and orthogonality matches closely how the PCA embedding is chosen. Alongside providing an intuitive understanding, we justify the statement with full theoretical analysis as well as with experiments.