CVPR 2016 Papers

Learned confidence measures gain increasing importance for outlier removal and quality improvement in stereo vision. However, acquiring the necessary training data is typically a tedious and time consuming task that involves manual interaction, active sensing devices and/or synthetic scenes. To overcome this problem, we propose a new, flexible, and scalable way for generating training data that only requires a set of stereo images as input. The key idea of our approach is to use different view points for reasoning about contradictions and consistencies between multiple depth maps generated with the same stereo algorithm. This enables us to generate a huge amount of training data in a fully automated manner. Among other experiments, we demonstrate the potential of our approach by boosting the performance of three learned confidence measures on the KITTI2012 dataset by simply training them on a vast amount of automatically generated training data rather than a limited amount of laser ground truth data.

Attributes offer useful mid-level features to interpret visual data. While most attribute learning methods are supervised by costly human-generated labels, we introduce a simple yet powerful unsupervised approach to learn and predict visual attributes directly from data. Given a large unlabeled image collection as input, we train deep Convolutional Neural Networks (CNNs) to output a set of discriminative, binary attributes often with semantic meanings. Specifically, we first train a CNN coupled with unsupervised discriminative clustering, and then use the cluster membership as a soft supervision to discover shared attributes from the clusters while maximizing their separability. The learned attributes are shown to be capable of encoding rich imagery properties from both natural images and contour patches. The visual representations learned in this way are also transferrable to other tasks such as object detection. We show other convincing results on the related tasks of image retrieval and classification, and contour detection.

Most existing person re-identification (Re-ID) approaches follow a supervised learning framework, in which a large number of labelled matching pairs are required for training. This severely limits their scalability in real-world applications. To overcome this limitation, we develop a novel cross-dataset transfer learning approach to learn a discriminative representation. It is unsupervised in the sense that the target dataset is completely unlabelled. Specifically, we present an multi-task dictionary learning method which is able to learn a dataset-shared but target-data-biased representation. Experimental results on five benchmark datasets demonstrate that the method significantly outperforms the state-of-the-art.

Scene understanding is a prerequisite to many high level tasks for any automated intelligent machine operating in real world environments. Recent attempts with supervised learning have shown promise in this direction but also highlighted the need for enormous quantity of supervised data --- performance increases in proportion to the amount of data used. However, this quickly becomes prohibitive when considering the manual labour needed to collect such data. In this work, we focus our attention on depth based semantic per-pixel labelling as a scene understanding problem and show the potential of computer graphics to generate virtually unlimited labelled data from synthetic 3D scenes. By carefully synthesizing training data with appropriate noise models we show comparable performance to state-of-the-art RGBD systems on NYUv2 dataset despite using only depth data as input and set a benchmark on depth-based segmentation on SUN RGB-D dataset.

We present a practical approach to address the problem of unconstrained face alignment for a single image. In our unconstrained problem, we need to deal with large shape and appearance variations under extreme head poses and rich shape deformation. To equip cascaded regressors with the capability to handle global shape variation and irregular appearance-shape relation in the unconstrained scenario, we partition the optimisation space into multiple domains of homogeneous descent, and predict a shape as a composition of estimations from multiple domain-specific regressors. With a specially formulated learning objective and a novel tree splitting function, our approach is capable of estimating a robust and meaningful composition. In addition to achieving state-of-the-art accuracy over existing approaches, our framework is also an efficient solution (350 FPS), thanks to the on-the-fly domain exclusion mechanism and the capability of leveraging the fast pixel feature.

In recent years, the task of estimating the 6D pose of object instances and complete scenes, i.e. camera localization, from a single input image has received considerable attention. Consumer RGB-D cameras have made this feasible, even for difficult, texture-less objects and scenes. In this work, we show that a single RGB image is sufficient to achieve visually convincing results. Our key concept is to model and exploit the uncertainty of the system at all stages of the processing pipeline. The uncertainty comes in the form of continuous distributions over 3D object coordinates and discrete distributions over object labels. We give three technical contributions. Firstly, we develop a regularized, auto-context regression framework which iteratively reduces uncertainty in object coordinate and object label predictions. Secondly, we introduce an efficient way to marginalize object coordinate distributions over depth. This is necessary to deal with missing depth information. Thirdly, we utilize the distributions over object labels to detect multiple objects simultaneously with a fixed budget of RANSAC hypotheses. We tested our system for object pose estimation and camera localization on commonly used data sets. We see a major improvement over competing systems.

The uncalibrated photometric stereo problem for non-Lambertian surfaces is challenging because of the large number of unknowns and its ill-posed nature stemming from unknown reflectance functions. We propose a model that represents various isotropic reflectance functions by using the principal components of items in a dataset, and formulate the uncalibrated photometric stereo as a regression problem. We then solve it by stepwise optimization utilizing principal components in order of their importance. We have also developed two techniques that lead to convergence and highly accurate reconstruction, namely (1) a coarse-to-fine approach with normal grouping, and (2) a randomized multipoint search. Our experimental results with synthetic data showed that our method significantly outperformed previous methods. We also evaluated the algorithm in terms of real image data, where it gave good reconstruction results.

3D shape recovery using photometric stereo (PS) gained increasing attention in the computer vision community in the last three decades due to its ability to recover the thinnest geometric structures. Yet, the reliabiliy of PS for color images is difficult to guarantee, because existing methods are usually formulated as the sequential estimation of the colored albedos, the normals and the depth. Hence, the overall reliability depends on that of each subtask. In this work we propose a new formulation of color photometric stereo, based on image ratios, that makes the technique independent from the albedos. This allows the unbiased 3D-reconstruction of colored surfaces in a single step, by solving a system of linear PDEs using a variational approach.

We present a real-time body orientation estimation in a micro-Unmanned Air Vehicle video stream. This work is part of a fully autonomous UAV system which can maneuver to face a single individual in challenging outdoor environments. Our body orientation estimation consists of the following steps: (a) obtaining a set of visual appearance models for each body orientation, where each model is tagged with a set of scene information (obtained from sensors); (b) exploiting the mutual information of on-board sensors using latent-dynamic conditional random fields (LDCRF); (c) Characterizing each visual appearance model with the most discriminative sensor information; (d) fast estimation of body orientation during the test flights given the LDCRF parameters and the corresponding sensor readings. The key aspects of our approach is to add sparsity to the sensor readings with latent variables followed by long range dependency analysis. Experimental results obtained over real-time video streams demonstrate a significant improvement in both speed (15-fps) and accuracy (72%) compared to the state of the art techniques that only rely on visual data. Video demonstration of our autonomous flights (both from ground view and aerial view) are included in the supplementary material.

This paper examines the problem of white-balance correction when a scene contains two illuminations. This is a two step process: 1) estimate the two illuminants; and 2) correct the image. Existing methods attempt to estimate a spatially varying illumination map, however, results are error prone and the resulting illumination maps are too low-resolution to be used for proper spatially varying white-balance correction. In addition, the spatially varying nature of these methods make them computationally intensive. We show that this problem can be effectively addressed by not attempting to obtain a spatially varying illumination map, but instead by performing illumination estimation on large sub-regions of the image. Our approach is able to detect when distinct illuminations are present in the image and accurately measure these illuminants. Since our proposed strategy is not suitable for spatially varying image correction, a user study is performed to see if there is a preference for how the image should be corrected when two illuminants are present, but only a global correction can be applied. The user study shows that when the illuminations are distinct, there is a preference for the outdoor illumination to be corrected resulting in warmer final result. We use these collective findings to demonstrate an effective two illuminant estimation scheme that produces corrected images that users prefer.

In this paper we present a system for performing low rank matrix factorization. Low-rank matrix factorization is an essential problem in many areas including computer vision, with applications in e.g. affine structure-from-motion, photometric stereo, and non-rigid structure from motion. We specifically target structured data patterns, with outliers and large amounts of missing data. Using recently developed characterizations of minimal solutions to matrix factorization problems with missing data, we show how these can be used as building blocks in a hierarchical system that performs bootstrapping on all levels. This gives an robust and fast system, with state-of-the-art performance.

Although promising results have been achieved in the areas of traffic-sign detection and classification, few works have provided simultaneous solutions to these two tasks for realistic real world images. We make two contributions to this problem. Firstly, we have created a large traffic-sign benchmark from 100000 Tencent Street View panoramas, going beyond previous benchmarks. It provides 100000 images containing 30000 traffic-sign instances. These images cover large variations in illuminance and weather conditions. Each traffic-sign in the benchmark is annotated with a class label, its bounding box and pixel mask. We call this benchmark Tsinghua-Tencent 100K. Secondly, we demonstrate how a robust end-to-end convolutional neural network (CNN) can simultaneously detect and classify traffic-signs. Most previous CNN image processing solutions target objects that occupy a large proportion of an image, and such networks do not work well for target objects occupying only a small fraction of an image like the traffic-signs here. Experimental results show the robustness of our network and its superiority to alternatives. The benchmark, source code and the CNN model introduced in this paper is publicly available.

The status quo approach to training object detectors requires expensive bounding box annotations. Our framework takes a markedly different direction: we transfer tracked object boxes from weakly-labeled videos to weakly-labeled images to automatically generate pseudo ground-truth boxes, which replace manually annotated bounding boxes. We first mine discriminative regions in the weakly-labeled image collection that frequently/rarely appear in the positive/negative images. We then match those regions to videos and retrieve the corresponding tracked object boxes. Finally, we design a hough transform algorithm to vote for the best box to serve as the pseudo GT for each image, and use them to train an object detector. Together, these lead to state-of-the-art weakly-supervised detection results on the PASCAL 2007 and 2010 datasets.

This paper presents an algorithm that allows to learn low dimensional representations of images in an unsupervised manner. The core idea is to combine two criteria that play important roles in unsupervised representation learning, namely sparsity and trace quotient. The former is known to be a convenient tool to identify underlying factors, and the latter is known as a disentanglement of underlying discriminative factors. In this work, we develop a generic cost function for learning jointly a sparsifying dictionary and a dimensionality reduction transformation. It leads to several counterparts of classic low dimensional representation methods, such as Principal Component Analysis, Local Linear Embedding, and Laplacian Eigenmap. Our proposed optimisation algorithm leverages the efficiency of geometric optimisation on Riemannian manifolds and a closed form solution to the elastic net problem.

Most existing person re-identification (re-id) models focus on matching still person images across disjoint camera views using the setting of either single-shot or multi-shot. Since limited information can be exploited from still images, it is hard (if not impossible) to overcome the occlusion, pose and camera-view change, and lighting variation problems. In comparison, video-based re-id methods can utilize extra space-time information, which contains much more rich cues for matching to overcome the mentioned problems. However, in this work, we find that when using video-based representation, some inter-class difference can be much more obscure than the one when using still-image-based representation, because different people could not only have similar appearance but also may have similar motions and actions which are hard to align. To solve this problem, we propose a top-push distance learning model (TDL), in which we integrate a top-push constrain, for matching video features of persons. The top-push constraint enforces the optimization on top-rank matching in re-id, so as to make the matching model more effective towards selecting more discriminative features to distinguish different persons. Our experiments show that the proposed video-based re-id framework outperforms the state-of-the-art video-based re-id methods.

Complex event detection on unconstrained Internet videos has seen much progress in recent years. However, state-of-the-art performance degrades dramatically when the number of positive training exemplars falls short. Since label acquisition is costly, laborious, and time-consuming, there is a real need to consider the much more challenging semantic event search problem, where no example video is given. In this paper, we present a state-of-the-art event search system without any example videos. Relying on the key observation that events (e.g. dog show) are usually compositions of multiple mid-level concepts (e.g. "dog," "theater," and "dog jumping"), we first train a skip-gram model to measure the relevance of each concept with the event of interest. The relevant concept classifiers then cast votes on the test videos but their reliability, due to lack of labeled training videos, has been largely unaddressed. We propose to combine the concept classifiers based on a principled estimate of their accuracy on the unlabeled test videos. A novel warping technique is proposed to improve the performance and an efficient highly-scalable algorithm is provided to quickly solve the resulting optimization. We conduct extensive experiments on the latest TRECVID MEDTest 2014, MEDTest 2013 and CCV datasets, and achieve state-of-the-art performances.

To image in high resolution large and occlusion-prone scenes, a camera must move above and around. Degradation of visibility due to geometric occlusions and distances is exacerbated by scattering, when the scene is in a participating medium. Moreover, underwater and in other media, artificial lighting is needed. Overall, data quality depends on the observed surface, medium and the time-varying poses of the camera and light source. This work proposes to optimize camera and light poses as they move, so that the surface is scanned efficiently and the descattered recovery has the highest quality. The work generalizes the next best view concept of robot vision to scattering media and cooperative movable lighting. It also extends descattering to platforms that move optimally. The optimization criterion is information gain, taken from information theory. We exploit the existence of a prior rough 3D model, since underwater such a model is routinely obtained using sonar. We demonstrate this principle in a scaled-down setup.

Recent face recognition experiments on a major benchmark LFW show stunning performance--a number of algorithms achieve near to perfect score, surpassing human recognition rates. In this paper, we advocate evaluations at the million scale (LFW includes only 13K photos of 5K people). To this end, we have assembled the MegaFace dataset and created the first MegaFace challenge. Our dataset includes One Million photos that capture more than 690K different individuals. The challenge evaluates performance of algorithms with increasing numbers of "distractors" (going from 10 to 1M) in the gallery set. We present both identification and verification performance, evaluate performance with respect to pose and a person's age, and compare as a function of training data size (#photos and #people). We report results of state of the art and baseline algorithms. The MegaFace dataset, baseline code, and evaluation scripts, are all publicly released for further experimentations at http://megaface.cs.washington.edu.

This paper studies the Tensor Robust Principal Component (TRPCA) problem which extends the known Robust PCA to the tensor case. Our model is based on a new tensor Singular Value Decomposition (t-SVD) and its induced tensor tubal rank and tensor nuclear norm. Consider that we have a 3-way tensor X in R^n*n*n_3 such that X=L_0+S_0, where L_0 has low tubal rank and S_0 is sparse. Is that possible to recover both components? In this work, we prove that under certain suitable assumptions, we can recover both the low-rank and the sparse components exactly by simply solving a convex program whose objective is a weighted combination of the tensor nuclear norm and the l1-norm, i.e., min L,E s.t. ||L||_*+lambda||E||_1 s.t. X=L+E. where lambda=1/sqrtmax(n_1,n_2)n_3. Interestingly, TRPCA involves RPCA as a special case when n_3=1 and thus it is a simple and elegant tensor extension of RPCA. Also numerical experiments verify our theory and the application for the image denoising demonstrates the effectiveness of our method.