CVPR 2016 Papers

This paper introduces situation recognition, the problem of producing a concise summary of the situation an image depicts including: (1) the main activity (e.g., clipping), (2) the participating actors, objects, substances, and locations (e.g., man, shears, sheep, wool, and field) and most importantly (3) the roles these participants play in the activity (e.g., the man is clipping, the shears are his tool, the wool is being clipped from the sheep, and the clipping is in a field). We use FrameNet, a verb and role lexicon developed by linguists, to define a large space of possible situations and collect a large-scale dataset containing over 500 activities, 1,700 roles, 11,000 objects, 125,000 images, and 200,000 unique situations. We also introduce structured prediction baselines and show that, in activity-centric images, situation-driven prediction of objects and activities outperforms independent object and activity recognition.

This paper aims to develop a method that can accurately estimate the crowd count from an individual image with arbitrary crowd density and arbitrary perspective. To this end,we have proposed a simple but effective Multi-column Convolutional Neural Network (MCNN) architecture to map the image to its crowd density map. The proposed MCNN allows the input image to be of arbitrary size or resolution. By utilizing filters with receptive fields of different sizes, the features learned by each column CNN are adaptive to variations in people/head size due to perspective effect or image resolution. Furthermore, the true density map is computed accurately based on geometry-adaptive kernels which do not need knowing the perspective map of the input image. Since exiting crowd counting datasets do not adequately cover all the challenging situations considered in our work, we have collected and labelled a large new dataset that includes 1198 images with about 330,000 heads annotated. On this challenging new dataset, as well as all existing datasets, we conduct extensive experiments to verify the effectiveness of the proposed model and method. In particular, with the proposed simple MCNN model, our method outperforms all existing methods. In addition, experiments show that our model, once trained on one dataset, can be readily transferred to a new dataset.

A methodology for 3D surface modeling from a single image is proposed. The principal novelty is concave and specular surface modeling without any externally imposed prior. The main idea of the method is to use BRDFs and generated rendered surfaces, to transfer the normal field, computed for the generated samples, to the unknown surface. The transferred information is adequate to blow and sculpt the segmented image mask in to a bas-relief of the object. The object surface is further refined basing on a photo-consistency formulation that relates for error minimization the original image and the modeled object.

We consider the problem of camera pose estimation for a scenario where the camera may have continuous and unknown changes in its focal length. Understanding frame by frame changes in camera focal length is vital to accurately estimating camera pose and vital to accurately render virtual objects in a scene with the correct perspective. However, most approaches to camera calibration require geometric constraints from many frames or the observation of a 3D calibration object --- both of which may not be feasible in augmented reality settings. This paper introduces a calibration objects based on a flat lenticular array that creates a color coded light-field whose observed color changes depending on the angle from which it is viewed. We derive an approach to estimate the focal length of the camera and the relative pose of an object from a single image. We characterize the performance of camera calibration across various focal lengths and camera models, and we demonstrate the advantages of the focal length estimation in rendering a virtual object in a video with constant zooming.

Event cameras are bio-inspired vision sensors which mimic retinas to measure per-pixel intensity change rather than outputting an actual intensity image. This proposed paradigm shift away from traditional frame cameras offers significant potential advantages: namely avoiding high data rates, dynamic range limitations and motion blur. Unfortunately, however, established computer vision algorithms may not at all be applied directly to event cameras. Methods proposed so far to reconstruct images, estimate optical flow, track a camera and reconstruct a scene come with severe restrictions on the environment or on the motion of the camera, e.g. allowing only rotation. Here, we propose, to the best of our knowledge, the first algorithm to simultaneously recover the motion field and brightness image, while the camera undergoes a generic motion through any scene. Our approach employs minimisation of a cost function that contains the asynchronous event data as well as spatial and temporal regularisation within a sliding window time interval. Our implementation relies on GPU-based optimisation and runs in near real-time. In a series of examples, we demonstrate the successful operation of our framework, including in situations where conventional cameras heavily suffer from dynamic range limitations or motion blur.

Near-Infrared (NIR) images of most materials exhibit less texture or albedo variations making them beneficial for vision tasks such as intrinsic image decomposition and structured light depth estimation. Understanding the reflectance properties (BRDF) of materials in the NIR wavelength range can be further useful for many photometric methods including shape from shading and inverse rendering. However, even with less albedo variation, many materials e.g. fabrics, leaves, etc. exhibit complex fine-scale surface detail making it hard to accurately estimate BRDF. In this paper, we present an approach to simultaneously estimate NIR BRDF and fine-scale surface details by imaging materials under different IR lighting and viewing directions. This is achieved by an iterative scheme that alternately estimates surface detail and NIR BRDF of materials. Our setup does not require complicated gantries or calibration and we present the first NIR dataset of 100 materials including a variety of fabrics (knits, weaves, cotton, satin, leather), and organic (skin, leaves, jute, trunk, fur) and inorganic materials (plastic, concrete, carpet). The NIR BRDFs measured from material samples are used with a shape-from-shading algorithm to demonstrate fine-scale reconstruction of objects from a single NIR image.

Estimating the number of clusters remains a difficult model selection problem. We consider this problem in the domain where the affinity relations involve groups of more than two nodes. Building on the previous formulation for the pairwise affinity case, we exploit the mathematical structures in the higher order case. We express the original minimal-rank and positive semi-definite (PSD) constraints in a form amenable for numerical implementation, as the original constraints are either intractable or even undefined in general in the higher order case. To scale to large problem sizes, we also propose an alternative formulation, so that it can be efficiently solved via stochastic optimization in an online fashion. We evaluate our algorithm with different applications to demonstrate its superiority, and show it can adapt to varying levels of unbalancedness of clusters.

In this paper, we introduce a similarity metric for curved shapes that can be described, distinctively, by ordered points. The proposed method represents a given curve as a point in the deformation space, the direct product of rigid transformation matrices, such that the successive action of the matrices on a fixed starting point reconstructs the full curve. In general, both open and closed curves are represented in the deformation space modulo shape orientation and orientation preserving diffeomorphisms. The use of direct product Lie groups to represent curved shapes led to an explicit formula for geodesic curves and the formulation of a similarity metric between shapes by the L2-norm on the Lie algebra. Additionally, invariance to reparametrization or estimation of point correspondence between shapes is performed as an intermediate step for computing geodesics. Furthermore, since there is no computation of differential quantities on the curves, our representation is more robust to local perturbations and needs no pre-smoothing. We compare our method with the elastic shape metric defined through the square root velocity (SRV) mapping, and other shape matching approaches.

Pose variation remains one of the major factors that adversely affect the accuracy of person re-identification. Such variation is not arbitrary as body parts (e.g. head, torso, legs) have relative stable spatial distribution. Breaking down the variability of global appearance regarding the spatial distribution potentially benefits the person matching. We therefore learn a novel similarity function, which consists of multiple sub-similarity measurements with each taking in charge of a subregion. In particular, we take advantage of the recently proposed polynomial feature map to describe the matching within each subregion, and inject all the feature maps into a unified framework. The framework not only outputs similarity measurements for different regions, but also makes a better consistency among them. Our framework can collaborate local similarities as well as global similarity to exploit their complementary strength. It is flexible to incorporate multiple visual cues to further elevate the performance. In experiments, we analyze the effectiveness of the major components. The results on four datasets show significant and consistent improvements over the state-of-the-art methods.

In this paper we present a tracker, which is radically different from state-of-the-art trackers: we apply no model updating, no occlusion detection, no combination of trackers, no geometric matching, and still deliver state-of-the-art tracking performance, as demonstrated on the popular online tracking benchmark (OTB) and six very challenging YouTube videos. The presented tracker simply matches the initial patch of the target in the first frame with candidates in a new frame and returns the most similar patch by a learned matching function. The strength of the matching function comes from being extensively trained generically, i.e., without any data of the target, using a Siamese deep neural network, which we design for tracking. Once learned, the matching function is used as is, without any adapting, to track previously unseen targets. It turns out that the learned matching function is so powerful that a simple tracker built upon it, coined Siamese INstance search Tracker, SINT, which only uses the original observation of the target from the first frame, suffices to reach state-of-the-art performance. Further, we show the proposed tracker even allows for target re-identification after the target was absent for a complete video shot.

In this paper, we introduce a novel shortlist computation algorithm for approximate, high-dimensional nearest neighbor search. Our method relies on a novel distance estimator: the residual-aware distance estimator, that accounts for the residual distances of data points to their respective quantized centroids, and uses it for accurate shortlist computation. Furthermore, we perform the residual-aware distance estimation with little additional memory and computational cost through simple pre-computation methods for inverted index and multi-index schemes. Because it modifies the initial shortlist collection phase, our new algorithm is applicable to most inverted indexing methods that use vector quantization. We have tested the proposed method with the inverted index and multi-index on a diverse set of benchmarks including up to one billion data points with varying dimensions, and found that our method robustly improves the accuracy of shortlists (up to 127% relatively higher) over the state-of-the-art techniques with a comparable or even faster computational cost.

Shape space is an active research field in computer vision study. The shape distance defined in a shape space may provide a simple and refined index to represent a unique shape. Wasserstein distance defines a Riemannian metric for the Wasserstein space. It intrinsically measures the similarities between shapes and is robust to image noise. Thus it has the potential for the 3D shape indexing and classification research. While the algorithms for computing Wasserstein distance have been extensively studied, most of them only work for genus-0 surfaces. This paper proposes a novel framework to compute Wasserstein distance between general topological surfaces with hyperbolic metric. The computational algorithms are based on Ricci flow, hyperbolic harmonic map, and hyperbolic power Voronoi diagram and the method is general and robust. We apply our method to study human facial expression, longitudinal brain cortical morphometry with normal aging, and cortical shape classification in Alzheimer's disease (AD). Experimental results demonstrate that our method may be used as an effective shape index, which outperforms some other standard shape measures in our AD versus healthy control classification study.

The prediction of salient areas in images has been traditionally addressed with hand-crafted features based on neuroscience principles. This paper, however, addresses the problem with a completely data-driven approach by training a convolutional neural network (convnet). The learning process is formulated as a minimization of a loss function that measures the Euclidean distance of the predicted saliency map with the provided ground truth. The recent publication of large datasets of saliency prediction has provided enough data to train end-to-end architectures that are both fast and accurate. Two designs are proposed: a shallow convnet trained from scratch, and a another deeper solution whose first three layers are adapted from another network trained for classification. To the authors knowledge, these are the first end-to-end CNNs trained and tested for the purpose of saliency prediction.

In this paper we present seven techniques that everybody should know to improve example-based single image super resolution (SR): 1) augmentation of data, 2) use of large dictionaries with efficient search structures, 3) cascading, 4) image self-similarities, 5) back projection refinement, 6) enhanced prediction by consistency check, and 7) context reasoning. We validate our seven techniques on standard SR benchmarks (i.e. Set5, Set14, B100) and methods (i.e. A+, SRCNN, ANR, Zeyde, Yang) and achieve substantial improvements.The techniques are widely applicable and require no changes or only minor adjustments of the SR methods. Moreover, our Improved A+ (IA) method sets new state-of-the-art results outperforming A+ by up to 0.9dB on average PSNR whilst maintaining a low time complexity.

Supervised contour detection methods usually require many labeled training images to obtain satisfactory performance. However, a large set of annotated data might be unavailable or extremely labor intensive. In this paper, we investigate the usage of semi-supervised learning (SSL) to obtain competitive detection accuracy with very limited training data (three labeled images). Specifically, we propose a semi-supervised structured ensemble learning approach for contour detection built on structured random forests (SRF). To allow SRF to be applicable to unlabeled data, we present an effective sparse representation approach to capture inherent structure in image patches by finding a compact and discriminative low-dimensional subspace representation in an unsupervised manner, enabling the incorporation of abundant unlabeled patches with their estimated structured labels to help SRF perform better node splitting. We re-examine the role of sparsity and propose a novel and fast sparse coding algorithm to boost the overall learning efficiency. To the best of our knowledge, this is the first attempt to apply SSL for contour detection. Extensive experiments on the BSDS500 segmentation dataset and the NYU Depth dataset demonstrate the superiority of the proposed method.

Despite significant progress in object categorization, in recent years, a number of important challenges remain; mainly, ability to learn from limited labeled data and ability to recognize object classes within large, potentially open, set of labels. Zero-shot learning is one way of addressing these challenges, but it has only been shown to work with limited sized class vocabularies and typically requires separation between supervised and unsupervised classes, allowing former to inform the latter but not vice versa. We propose the notion of semi-supervised vocabulary-informed learning to alleviate the above mentioned challenges and address problems of supervised, zero-shot and open set recognition using a unified framework. Specifically, we propose a maximum margin framework for semantic manifold-based recognition that incorporates distance constraints from (both supervised and unsupervised) vocabulary atoms, ensuring that labeled samples are projected closest to their correct prototypes, in the embedding space, than to others. We show that resulting model shows improvements in supervised, zero-shot, and large open set recognition, with up to 310K class vocabulary on AwA and ImageNet datasets.

The state-of-the-art in semantic segmentation is currently represented by fully convolutional networks (FCNs). However, FCNs use large receptive fields and many pooling layers, both of which cause blurring and low spatial resolution in the deep layers. As a result FCNs tend to produce segmentations that are poorly localized around object boundaries. Prior work has attempted to address this issue in post-processing steps, for example using a color-based CRF on top of the FCN predictions. However, these approaches require additional parameters and low-level features that are difficult to tune and integrate into the original network architecture. Additionally, most CRFs use color-based pixel affinities, which are not well suited for semantic segmentation and lead to spatially disjoint predictions. To overcome these problems, we introduce a Boundary Neural Field (BNF), which is a global energy model integrating FCN predictions with boundary cues. The boundary information is used to enhance semantic segment coherence and to improve object localization. Specifically, we first show that the convolutional filters of semantic FCNs provide good features for boundary detection. We then employ the predicted boundaries to define pairwise potentials in our energy. Finally, we show that our energy decomposes semantic segmentation into multiple binary problems, which can be relaxed for efficient global optimization. We report extensive experiments demonstrating that minimization of our global boundary-based energy yields results superior to prior globalization methods, both quantitatively as well as qualitatively.

Semantic object parsing is a fundamental task for understanding objects in detail in computer vision community, where incorporating multi-level contextual information is critical for achieving such fine-grained pixel-level recognition. Prior methods often leverage the contextual information through post-processing predicted confidence maps. In this work, we propose a novel deep Local-Global Long Short-Term Memory (LG-LSTM) architecture to seamlessly incorporate short-distance and long-distance spatial dependencies into the feature learning over all pixel positions. In each LG-LSTM layer, local guidance from neighboring positions and global guidance from the whole image are imposed on each position to better exploit complex local and global contextual information. Individual LSTMs for distinct spatial dimensions are also utilized to intrinsically capture various spatial layouts of semantic parts in the images, yielding distinct hidden and memory cells of each position for each dimension. In our parsing approach, several LG-LSTM layers are stacked and appended to the intermediate convolutional layers to directly enhance visual features, allowing network parameters to be learned in an end-to-end way. The long chains of sequential computation by stacked LG-LSTM layers also enable each pixel to sense a much larger region for inference benefiting from the memorization of previous dependencies in all positions along all dimensions. Comprehensive evaluations on three public datasets well demonstrate the significant superiority of our LG-LSTM over other state-of-the-art methods for object parsing.

This supplementary material provides additional illustrations, visualizations and experiments. We start by showing the color coding and label mapping used for the semantic and instance label results in the paper. Then we provide more details about the 3D fold/curb detection and parameter settings that are used in the paper. Next, we provide additional quantitative and qualitative semi-dense inference results for both semantic and instance segmentation. Finally, we show the ability of our method to annotate 3D point clouds with semantic and instance labels which is a byproduct of our approach.