DensePose provides a pixel-accurate association of images with 3D mesh coordinates, but does not provide a 3D mesh, while Human Mesh Reconstruction (HMR) systems have high 2D reprojection error, as measured by DensePose localization metrics. In this work we introduce MeshPose to jointly tackle DensePose and HMR. For this we first introduce new losses that allow us to use weak DensePose supervision to accurately localize in 2D a subset of the mesh vertices (‘VertexPose’). We then lift these vertices to 3D, yielding a low-poly body mesh (‘MeshPose’). Our system is trained in an end -to-end manner and is the first HMR method to attain competitive DensePose accuracy, while also being lightweight and amenable to efficient inference, making it suitable for real-time AR applications.

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.

In this work, we tackle the problem of instance segmentation, the task of simultaneously solving object detection and semantic segmentation. Towards this goal, we present a model, called MaskLab, which produces three outputs: box detection, semantic segmentation, and direction prediction. Building on top of the Faster-RCNN object detector, the predicted boxes provide accurate localization of object instances. Within each region of interest, MaskLab performs foreground/background segmentation by combining semantic and direction prediction. Semantic segmentation assists the model in distinguishing between objects of different semantic classes including background, while the direction prediction, estimating each pixel's direction towards its corresponding center, allows separating instances of the same semantic class. Moreover, we explore the effect of incorporating recent successful methods from both segmentation and detection (i.e. atrous convolution and hypercolumn). Our proposed model is evaluated on the COCO instance segmentation benchmark and shows comparable performance with other state-of-art models.

The design of neural network architectures is an important component for achieving state-of-the-art performance with machine learning systems across a broad array of tasks. Much work has endeavored to design and build architectures automatically through clever construction of a search space paired with simple learning algorithms. Recent progress has demonstrated that such meta-learning methods may exceed scalable human-invented architectures on image classification tasks. An open question is the degree to which such methods may generalize to new domains. In this work we explore the construction of meta-learning techniques for dense image prediction focused on the tasks of scene parsing, person-part segmentation, and semantic image segmentation. Constructing viable search spaces in this domain is challenging because of the multi-scale representation of visual information and the necessity to operate on high resolution imagery. Based on a survey of techniques in dense image prediction, we construct a recursive search space and demonstrate that even with efficient random search, we can identify architectures that outperform human-invented architectures and achieve state-of-the-art performance on three dense prediction tasks including 82.7% on Cityscapes (street scene parsing), 71.3% on PASCAL-Person-Part (person-part segmentation), and 87.9% on PASCAL VOC 2012 (semantic image segmentation). Additionally, the resulting architecture is more computationally efficient, requiring half the parameters and half the computational cost as previous state of the art systems.

In this work we address the task of semantic image segmentation with Deep Learning and make three main contributions that are experimentally shown to have substantial practical merit. First, we highlight convolution with upsampled filters, or 'atrous convolution', as a powerful tool in dense prediction tasks. Atrous convolution allows us to explicitly control the resolution at which feature responses are computed within Deep Convolutional Neural Networks. It also allows us to effectively enlarge the field of view of filters to incorporate larger context without increasing the number of parameters or the amount of computation. Second, we propose atrous spatial pyramid pooling (ASPP) to robustly segment objects at multiple scales. ASPP probes an incoming convolutional feature layer with filters at multiple sampling rates and effective fields-of-views, thus capturing objects as well as image context at multiple scales. Third, we improve the localization of object boundaries by combining methods from DCNNs and probabilistic graphical models. The commonly deployed combination of max-pooling and downsampling in DCNNs achieves invariance but has a toll on localization accuracy. We overcome this by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF), which is shown both qualitatively and quantitatively to improve localization performance. Our proposed "DeepLab" system sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 79.7 percent mIOU in the test set, and advances the results on three other datasets: PASCAL-Context, PASCAL-Person-Part, and Cityscapes. All of our code is made publicly available online.

We propose a method for multi-person detection and 2-D pose estimation that achieves state-of-art results on the challenging COCO keypoints task. It is a simple, yet powerful, top-down approach consisting of two stages. In the first stage, we predict the location and scale of boxes which are likely to contain people, for this we use the Faster RCNN detector. In the second stage, we estimate the keypoints of the person potentially contained in each proposed bounding box. For each keypoint type we predict dense heatmaps and offsets using a fully convolutional ResNet. To combine these outputs we introduce a novel aggregation procedure to obtain highly localized keypoint predictions. We also use a novel form of keypoint-based Non-Maximum-Suppression (NMS), instead of the cruder box-level NMS, and a novel form of keypoint-based confidence score estimation, instead of box-level scoring. Trained on COCO data alone, our final system achieves average precision of 0.649 on the COCO test-dev set and the 0.643 test-standard sets, outperforming the winner of the 2016 COCO keypoints challenge and other recent state-of-art. Further, by using additional in-house labeled data we obtain an even higher average precision of 0.685 on the test-dev set and 0.673 on the test-standard set, more than 5% absolute improvement compared to the previous best performing method on the same dataset.

Deep convolutional neural networks (CNNs) are the backbone of state-of-art semantic image segmentation systems. Recent work has shown that complementing CNNs with fully-connected conditional random fields (CRFs) can significantly enhance their object localization accuracy, yet dense CRF inference is computationally expensive. We propose replacing the fully-connected CRF with domain transform (DT), a modern edge-preserving filtering method in which the amount of smoothing is controlled by a reference edge map. Domain transform filtering is several times faster than dense CRF inference and we show that it yields comparable semantic segmentation results, accurately capturing object boundaries. Importantly, our formulation allows learning the reference edge map from intermediate CNN features instead of using the image gradient magnitude as in standard DT filtering. This produces task-specific edges in an end-to-end trainable system optimizing the target semantic segmentation quality.

Deep Convolutional Neural Networks (DCNNs) achieve invariance to domain transformations (deformations) by using multiple `max-pooling' (MP) layers. In this work we show that alternative methods of modeling deformations can improve the accuracy and efficiency of DCNNs. First, we introduce epitomic convolution as an alternative to the common convolution-MP cascade of DCNNs, that comes with the same computational cost but favorable learning properties. Second, we introduce a Multiple Instance Learning algorithm to accommodate global translation and scaling in image classification, yielding an efficient algorithm that trains and tests a DCNN in a consistent manner. Third we develop a DCNN sliding window detector that explicitly, but efficiently, searches over the object's position, scale, and aspect ratio. We provide competitive image classification and localization results on the ImageNet dataset and object detection results on Pascal VOC2007.

We present a system which can recognize the contents of your meal from a single image, and then predict its nutritional contents, such as calories. The simplest version assumes that the user is eating at a restaurant for which we know the menu. In this case, we can collect images offline to train a multi-label classifier. At run time, we apply the classifier (running on your phone) to predict which foods are present in your meal, and we lookup the corresponding nutritional facts. We apply this method to a new dataset of images from 23 different restaurants, using a CNN-based classifier, significantly outperforming previous work. The more challenging setting works outside of restaurants. In this case, we need to estimate the size of the foods, as well as their labels. This requires solving segmentation and depth / volume estimation from a single image. We present CNN-based approaches to these problems, with promising preliminary results.

Deep convolutional neural networks (DCNNs) trained on a large number of images with strong pixel-level annotations have recently significantly pushed the state-of-art in semantic image segmentation. We study the more challenging problem of learning DCNNs for semantic image segmentation from either (1) weakly annotated training data such as bounding boxes or image-level labels or (2) a combination of few strongly labeled and many weakly labeled images, sourced from one or multiple datasets. We develop Expectation-Maximization (EM) methods for semantic image segmentation model training under these weakly supervised and semi-supervised settings. Extensive experimental evaluation shows that the proposed techniques can learn models delivering competitive results on the challenging PASCAL VOC 2012 image segmentation benchmark, while requiring significantly less annotation effort. We share source code implementing the proposed system at https://bitbucket.org/deeplab/deeplab-public.

The goal of this paper is to question the necessity of features like SIFT in categorical visual recognition tasks. As an alternative, we develop a generative model for the raw intensity of image patches and show that it can support image classification performance on par with optimized SIFT-based techniques in a bag-of-visual-words setting. Key ingredient of the proposed model is a compact dictionary of mini-epitomes, learned in an unsupervised fashion on a large collection of images. The use of epitomes allows us to explicitly account for photometric and position variability in image appearance. We show that this flexibility considerably increases the capacity of the dictionary to accurately approximate the appearance of image patches and support recognition tasks. For image classification, we develop histogram-based image encoding methods tailored to the epitomic representation, as well as an "epitomic footprint" encoding which is easy to visualize and highlights the generative nature of our model. We discuss in detail computational aspects and develop efficient algorithms to make the model scalable to large tasks. The proposed techniques are evaluated with experiments on the challenging PASCAL VOC 2007 image classification benchmark.

The first main contribution of this paper is a novel method for representing images based on a dictionary of shape epitomes. These shape epitomes represent the local edge structure of the image and include hidden variables to encode shift and rotations. They are learnt in an unsupervised manner from groundtruth edges. This dictionary is compact but is also able to capture the typical shapes of edges in natural images. In this paper, we illustrate the shape epitomes by applying them to the image labeling task. In other work, described in the supplementary material, we apply them to edge detection and image modeling. We apply shape epitomes to image labeling by using Conditional Random Field (CRF) Models. They are alternatives to the superpixel or pixel representations used in most CRFs. In our approach, the shape of an image patch is encoded by a shape epitome from the dictionary. Unlike the superpixel representation, our method avoids making early decisions which cannot be reversed. Our resulting hierarchical CRFs efficiently capture both local and global class co-occurrence properties. We demonstrate its quantitative and qualitative properties of our approach with image labeling experiments on two standard datasets: MSRC-21 and Stanford Background.

We propose a novel way to induce a random field from an energy function on discrete labels. It amounts to locally injecting noise to the energy potentials, followed by finding the global minimum of the perturbed energy function. The resulting Perturb-and-MAP random fields harness the power of modern discrete energy minimization algorithms, effectively transforming them into efficient random sampling algorithms, thus extending their scope beyond the usual deterministic setting. In this fashion we can enjoy the benefits of a sound probabilistic framework, such as the ability to represent the solution uncertainty or learn model parameters from training data, while completely bypassing costly Markov-chain Monte-Carlo procedures typically associated with discrete label Gibbs Markov random fields (MRFs). We study some interesting theoretical properties of the proposed model in juxtaposition to those of Gibbs MRFs and address the issue of principled design of the perturbation process. We present experimental results in image segmentation and scene labeling that illustrate the new qualitative aspects and the potential of the proposed model for practical computer vision applications.

We present a technique for exact simulation of Gaussian Markov random fields (GMRFs), which can be interpreted as locally injecting noise to each Gaussian factor independently, followed by computing the mean/mode of the perturbed GMRF. Coupled with standard iterative techniques for the solution of symmetric positive definite systems, this yields a very efficient sampling algorithm with essentially linear complexity in terms of speed and memory requirements, well suited to extremely large scale probabilistic models. Apart from synthesizing data under a Gaussian model, the proposed technique directly leads to an efficient unbiased estimator of marginal variances. Beyond Gaussian models, the proposed algorithm is also very useful for handling highly non-Gaussian continuously-valued MRFs such as those arising in statistical image modeling or in the first layer of deep belief networks describing real-valued data, where the non-quadratic potentials coupling different sites can be represented as finite or infinite mixtures of Gaussians with the help of local or distributed latent mixture assignment variables. The Bayesian treatment of such models most naturally involves a block Gibbs sampler which alternately draws samples of the conditionally independent latent mixture assignments and the conditionally multivariate Gaussian continuous vector and we show that it can directly benefit from the proposed methods.

We present a novel multiscale image representation belonging to the class of multiscale multiplicative decompositions, which we term Poisson-Haar transform. The proposed representation is well-suited for analyzing images degraded by signal-dependent Poisson noise, allowing efficient estimation of their underlying intensity by means of multiscale Bayesian schemes. The Poisson-Haar decomposition has a direct link to the standard 2D Haar wavelet transform, thus retaining many of the properties that have made wavelets successful in signal processing and analysis. The practical relevance and effectiveness of the proposed approach is verified through denoising experiments on simulated and real-world photon-limited images.

We are interested in recovering aspects of vocal tract's geometry and dynamics from speech, a problem referred to as speech inversion. Traditional audio-only speech inversion techniques are inherently ill-posed since the same speech acoustics can be produced by multiple articulatory configurations. To alleviate the ill-posedness of the audio-only inversion process, we propose an inversion scheme which also exploits visual information from the speaker's face. The complex audiovisual-to-articulatory mapping is approximated by an adaptive piecewise linear model. Model switching is governed by a Markovian discrete process which captures articulatory dynamic information. Each constituent linear mapping is effectively estimated via canonical correlation analysis. In the described multimodal context, we investigate alternative fusion schemes which allow interaction between the audio and visual modalities at various synchronization levels. For facial analysis, we employ active appearance models (AAMs) and demonstrate fully automatic face tracking and visual feature extraction. Using the AAM features in conjunction with audio features such as Mel frequency cepstral coefficients (MFCCs) or line spectral frequencies (LSFs) leads to effective estimation of the trajectories followed by certain points of interest in the speech production system. We report experiments on the QSMT and MOCHA databases which contain audio, video, and electromagnetic articulography data recorded in parallel. The results show that exploiting both audio and visual modalities in a multistream hidden Markov model based scheme clearly improves performance relative to either audio or visual-only estimation.

While the accuracy of feature measurements heavily depends on changing environmental conditions, studying the consequences of this fact in pattern recognition tasks has received relatively little attention to date. In this paper, we explicitly take feature measurement uncertainty into account and show how multimodal classification and learning rules should be adjusted to compensate for its effects. Our approach is particularly fruitful in multimodal fusion scenarios, such as audiovisual speech recognition, where multiple streams of complementary time-evolving features are integrated. For such applications, provided that the measurement noise uncertainty for each feature stream can be estimated, the proposed framework leads to highly adaptive multimodal fusion rules which are easy and efficient to implement. Our technique is widely applicable and can be transparently integrated with either synchronous or asynchronous multimodal sequence integration architectures. We further show that multimodal fusion methods relying on stream weights can naturally emerge from our scheme under certain assumptions; this connection provides valuable insights into the adaptivity properties of our multimodal uncertainty compensation approach. We show how these ideas can be practically applied for audiovisual speech recognition. In this context, we propose improved techniques for person-independent visual feature extraction and uncertainty estimation with active appearance models, and also discuss how enhanced audio features along with their uncertainty estimates can be effectively computed. We demonstrate the efficacy of our approach in audiovisual speech recognition experiments on the CUAVE database using either synchronous or asynchronous multimodal integration models.

We present an improved statistical model for analyzing Poisson processes, with applications to photon-limited imaging. We build on previous work, adopting a multiscale representation of the Poisson process in which the ratios of the underlying Poisson intensities (rates) in adjacent scales are modeled as mixtures of conjugate parametric distributions. Our main contributions include: 1) a rigorous and robust regularized expectation-maximization (EM) algorithm for maximum-likelihood estimation of the rate-ratio density parameters directly from the noisy observed Poisson data (counts); 2) extension of the method to work under a multiscale hidden Markov tree model (HMT) which couples the mixture label assignments in consecutive scales, thus modeling interscale coefficient dependencies in the vicinity of image edges; 3) exploration of a 2-D recursive quad-tree image representation, involving Dirichlet-mixture rate-ratio densities, instead of the conventional separable binary-tree image representation involving beta-mixture rate-ratio densities; and 4) a novel multiscale image representation, which we term Poisson-Haar decomposition, that better models the image edge structure, thus yielding improved performance. Experimental results on standard images with artificially simulated Poisson noise and on real photon-limited images demonstrate the effectiveness of the proposed techniques.

Parametric models of shape and texture such as active appearance models (AAMs) are diverse tools for deformable object appearance modeling and have found important applications in both image synthesis and analysis problems. Among the numerous algorithms that have been proposed for AAM fitting, those based on the inverse-compositional image alignment technique have recently received considerable attention due to their potential for high efficiency. However, existing fitting algorithms perform poorly when used in conjunction with models exhibiting significant appearance variation, such as AAMs trained on multiple-subject human face images. We introduce two enhancements to inverse-compositional AAM matching algorithms in order to overcome this limitation. First, we propose fitting algorithm adaptation, by means of (a) fitting matrix adjustment and (b) AAM mean template update. Second, we show how prior information can be incorporated and constrain the AAM fitting process. The inverse-compositional nature of the algorithm allows efficient implementation of these enhancements. Both techniques substantially improve AAM fitting performance, as demonstrated with experiments on publicly available multi-face datasets.

We are interested in recovering aspects of vocal tract's geometry and dynamics from auditory and visual speech cues. We approach the problem in a statistical framework based on Hidden Markov Models and demonstrate effective estimation of the trajectories followed by certain points of interest in the speech production system. Alternative fusion schemes are investigated to account for asynchrony between the modalities and allow independent modeling of the dynamics of the involved streams. Visual cues are extracted from the speaker's face by means of active appearance modeling. We report experiments on the QSMT database which contains audio, video, and electromagnetic articulography data recorded in parallel. The results show that exploiting both audio and visual modalities in a multistream HMM based scheme clearly improves performance relative to either audio or visual-only estimation.

We present a novel technique for image inpainting, the problem of filling-in missing image parts. Image inpainting is ill-posed and we adopt a probabilistic model-based approach to regularize it. The main elements of our image model are, first, an over-complete complex-wavelet image representation, which ensures good shift invariance and directional selectivity and, second, a discrete-state/continuous-observation hidden Markov tree model for the wavelet coefficients, which captures key statistical properties of natural image wavelet responses, such as heavy-tailed histograms and persistence of large wavelet coefficients across scales. We show how these ideas can be integrated into a multi-scale generative process for natural images and present alternative deterministic and Markov chain Monte Carlo algorithms for image inpainting under this model. We demonstrate the effectiveness of the method in digitally restoring images of ancient wall-paintings.

We present an improved statistical model of Poisson processes, with applications to photon-limited imaging. We build on previous work, adopting a multiscale representation of the Poisson process in which the ratios of the underlying Poisson intensities (rates) in adjacent scales are modeled as mixtures of conjugate parametric distributions. Our main novel contributions are (1) a rigorous and robust regularized expectation-maximization (EM) algorithm for maximum-likelihood estimation of the rate-ratio density parameters directly from the observed Poisson data (counts); (2) extension of the method to work under a scale-recursive hidden Markov tree model (HMT) which couples the mixture label assignments in consecutive scales, thus modeling inter-scale coefficient dependencies in the vicinity of edges; and (3) exploration of a fully 2-D quad-tree image partitioning, involving Dirichlet-mixture rate-ratio densities, instead of the conventional separable binary image partitioning involving Beta-mixture rate-ratio densities. Experimental intensity estimation results on standard images with artificially simulated Poisson noise and photon-limited images with real shot noise demonstrate the effectiveness of the proposed approach.

We address the problem of audiovisual speech inversion, namely recovering the vocal tract's geometry from auditory and visual speech cues. We approach the problem in a statistical framework, combining ideas from multistream Hidden Markov Models and canonical correlation analysis, and demonstrate effective estimation of the trajectories followed by certain points of interest in the speech production system. Our experiments show that exploiting both audio and visual modalities clearly improves performance relative to either audio-only or visual-only estimation. We report experiments on the QSMT database which contains audio, video, and electromagnetic articulography data recorded in parallel.

We study the effect of uncertain feature measurements and show how classification and learning rules should be adjusted to compensate for it. Our approach is particularly fruitful in multimodal fusion scenarios, such as audio-visual speech recognition, where multiple streams of complementary features whose reliability is time-varying are integrated. For such applications, by taking the measurement noise uncertainty of each feature stream into account, the proposed framework leads to highly adaptive multimodal fusion rules for classification and learning which are widely applicable and easy to implement. We further show that previous multimodal fusion methods relying on stream weights fall under our scheme under certain assumptions; this provides novel insights into their applicability for various tasks and suggests new practical ways for estimating the stream weights adaptively. The potential of our approach is demonstrated in audio-visual speech recognition experiments.

Geometric active contour models are very popular partial differential equation-based tools in image analysis and computer vision. We present a new multigrid algorithm for the fast evolution of level-set-based geometric active contours and compare it with other established numerical schemes. We overcome the main bottleneck associated with most numerical implementations of geometric active contours, namely the need for very small time steps to avoid instability, by employing a very stable fully 2-D implicit-explicit time integration numerical scheme. The proposed scheme is more accurate and has improved rotational invariance properties compared with alternative split schemes, particularly when big time steps are utilized. We then apply properly designed multigrid methods to efficiently solve the occurring sparse linear system. The combined algorithm allows for the rapid evolution of the contour and convergence to its final configuration after very few iterations. Image segmentation experiments demonstrate the efficiency and accuracy of the method.

While the accuracy of feature measurements heavily depends on changing environmental conditions, studying the consequences of this fact in pattern recognition tasks has received relatively little attention to date. In this work we explicitly take into account feature measurement uncertainty and we show how classification rules should be adjusted to compensate for its effects. Our approach is particularly fruitful in multimodal fusion scenarios, such as audio-visual speech recognition, where multiple streams of complementary time-evolving features are integrated. For such applications, provided that the measurement noise uncertainty for each feature stream can be estimated, the proposed framework leads to highly adaptive multimodal fusion rules which are widely applicable and easy to implement. We further show that previous multimodal fusion methods relying on stream weights fall under our scheme under certain assumptions; this provides novel insights into their applicability for various tasks and suggests new practical ways for estimating the stream weights adaptively. The potential of our approach is demonstrated in audio-visual speech recognition using either synchronous or asynchronous models.

Scale-spaces induced by diffusion processes play an important role in many computer vision tasks. Automatically selecting the most appropriate scale for a particular problem is a central issue for the practical applicability of such scale-space techniques. This paper concentrates on automatic scale selection when nonlinear diffusion scale-spaces are utilized for image denoising. The problem is studied in a statistical model selection framework and cross-validation techniques are utilized to address it in a principled way. The proposed novel algorithms do not require knowledge of the noise variance and have acceptable computational cost. Extensive experiments on natural images show that the proposed methodology leads to robust algorithms, which outperform existing techniques for a wide range of noise types and noise levels.

Multiscale, i.e. scale-space image analysis is a powerful framework for many image processing tasks. A fundamental issue with such scale-space techniques is the automatic selection of the most salient scale for a particular application. This paper considers optimal scale selection when nonlinear diffusion and morphological scale-spaces are utilized for image denoising. The problem is studied from a statistical model selection viewpoint and cross-validation techniques are utilized to address it in a principled way. The proposed novel algorithms do not require knowledge of the noise variance, have acceptable computational cost and are readily integrated with a wide class of scale-space inducing processes which require setting of a scale parameter. Our experiments show that this methodology leads to robust algorithms, which outperform existing scale-selection techniques for a wide range of noise types and noise levels.