ICCV 2017 Papers

This paper focuses on the task of room layout estimation from a monocular RGB image. Prior works break the problem into two sub-tasks: semantic segmentation of floor, walls, ceiling to produce layout hypotheses, followed by an iterative optimization step to rank these hypotheses. In contrast, we adopt a more direct formulation of this problem as one of estimating an ordered set of room layout keypoints. The room layout and the corresponding segmentation is completely specified given the locations of these ordered keypoints. We predict the locations of the room layout keypoints using RoomNet, an end-to-end trainable encoder-decoder network. On the challenging benchmark datasets Hedau and LSUN, we achieve state-of-the-art performance along with 200x to 600x speedup compared to the most recent work. Additionally, we present optional extensions to the RoomNet architecture such as including recurrent computations and memory units to refine the keypoint locations under the same parametric capacity.

In this paper, we develop a modified differential Structure from Motion (SfM) algorithm that can estimate relative pose from two frames despite of Rolling Shutter (RS) artifacts. In particular, we show that under constant velocity assumption, the errors induced by the rolling shutter effect can be easily rectified by a linear scaling operation on each optical flow. We further propose a 9-point algorithm to recover the relative pose of a rolling shutter camera that undergoes constant acceleration motion. We demonstrate that the dense depth maps recovered from the relative pose of the RS camera can be used in a RS-aware warping for image rectification to recover high-quality Global Shutter (GS) images. Experiments on both synthetic and real RS images show that our RS-aware differential SfM algorithm produces more accurate results on relative pose estimation and 3D reconstruction from images distorted by RS effect compared to standard SfM algorithms that assume a GS camera model. We also demonstrate that our RS-aware warping for image rectification method outperforms state-of-the-art commercial software products, i.e. Adobe After Effects and Apple Imovie, at removing RS artifacts.

A vast majority of consumer cameras operate the rolling shutter mechanism, which often produces distorted images due to inter-row delay while capturing an image. Recent methods for monocular rolling shutter compensation utilize blur kernel, straightness of line segments, as well as angle and length preservation. However, they do not incorporate scene geometry explicitly for rolling shutter correction, therefore, information about the 3D scene geometry is often distorted by the correction process. In this paper we propose a novel method which leverages geometric properties of the scene---in particular vanishing directions---to estimate the camera motion during rolling shutter exposure from a single distorted image. The proposed method jointly estimates the orthogonal vanishing directions and the rolling shutter camera motion. We performed extensive experiments on synthetic and real datasets which demonstrate the benefits of our approach both in terms of qualitative and quantitative results (in terms of a geometric structure fitting) as well as with respect to computation time.

Video super-resolution (SR) aims to generate a high-resolution (HR) frame from multiple low-resolution (LR) frames. The inter-frame temporal relation is as crucial as the intra-frame spatial relation for tackling this problem. However, how to utilize temporal information efficiently and effectively remains challenging since complex motion is difficult to model and can introduce adverse effects if not handled properly. We address this problem from two aspects. First, we propose a temporal adaptive neural network that can adaptively determine the optimal scale of temporal dependency. Filters on various temporal scales are applied to the input LR sequence before their responses are adaptively aggregated. Second, we reduce the complexity of motion between neighboring frames using a spatial alignment network that is much more robust and efficient than competing alignment methods and can be jointly trained with the temporal adaptive network in an end-to-end manner. Our proposed models with learned temporal dynamics are systematically evaluated on public video datasets and achieve state-of-the-art SR results compared with other recent video SR approaches. Both of the temporal adaptation and the spatial alignment modules are demonstrated to considerably improve SR quality over their plain counterparts.

Markov Random Fields are widely used to model light-field stereo matching problems. However, most previous approaches used fixed parameters and did not adapt to light-field statistics. Instead, they explored explicit vision cues to provide local adaptability and thus enhanced depth quality. But such additional assumptions could end up confining their applicability, e.g. algorithms designed for dense light fields are not suitable for sparse ones. In this paper, we develop an empirical Bayesian framework--Robust Pseudo Random Field--to explore intrinsic statistical cues for broad applicability. Based on pseudo-likelihood, it applies soft expectation-maximization (EM) for good model fitting and hard EM for robust depth estimation. We introduce novel pixel difference models to enable such adaptability and robustness simultaneously. We also devise an algorithm to employ this framework on dense, sparse, and even denoised light fields. Experimental results show that it estimates scene-dependent parameters robustly and converges quickly. In terms of depth accuracy and computation speed, it also outperforms state-of-the-art algorithms constantly.

Recently deep neural networks have been widely employed to deal with the visual tracking problem. In this work, we present a new deep architecture which incorporates the temporal and spatial information to boost the tracking performance. Our deep architecture contains three networks, a Feature Net, a Temporal Net, and a Spatial Net. The Feature Net extracts general feature representations of the target. With these feature representations, the Temporal Net encodes the trajectory of the target and directly learns temporal correspondences to estimate the object state from a global perspective. Based on the learning results of the Temporal Net, the Spatial Net further refines the object tracking state using local spatial object information. Extensive experiments on four of the largest tracking benchmarks, including VOT2014, VOT2016, OTB50, and OTB100, demonstrate competing performance of the proposed tracker over a number of state-of-the-art algorithms.

Dictionary learning and component analysis are part of one of the most well-studied and active research fields, at the intersection of signal and image processing, computer vision, and statistical machine learning. In dictionary learning, the current methods of choice are arguably K-SVD and its variants, which learn a dictionary (i.e., a decomposition) for sparse coding via Singular Value Decomposition. In robust component analysis, leading methods derive from Principal Component Pursuit (PCP), which recovers a low-rank matrix from sparse corruptions of unknown magnitude and support. However, K-SVD is sensitive to the presence of noise and outliers in the training set. Additionally, PCP does not provide a dictionary that respects the structure of the data (e.g., images), and requires expensive SVD computations when solved by convex relaxation. In this paper, we introduce a new robust decomposition of images by combining ideas from sparse dictionary learning and PCP. We propose a novel Kronecker-decomposable component analysis which is robust to gross corruption, can be used for low-rank modeling, and leverages separability to solve significantly smaller problems. We design an efficient learning algorithm by drawing links with a restricted form of tensor factorization. The effectiveness of the proposed approach is demonstrated on real-world applications, namely background subtraction and image denoising, by performing a thorough comparison with the current state of the art.

This paper proposes a robust solution for accurate 3D hand pose estimation in the presence of an external object interacting with hands. Our main insight is that the shape of an object causes a configuration of the hand in the form of a hand grasp. Along this line, we simultaneously train deep neural networks using paired depth images. The object-oriented network learns functional grasps from an object perspective, whereas the hand-oriented network explores the details of hand configurations from a hand perspective. The two networks share intermediate observations produced from different perspectives to create a more informed representation. Our system then collaboratively classifies the grasp types and orientation of the hand and further constrains a pose space using these estimates. Finally, we collectively refine the unknown pose parameters to reconstruct the final hand pose. To this end, we conduct extensive evaluations to validate the efficacy of the proposed collaborative learning approach by comparing it with self-generated baselines and the state-of-the-art method.

The success of deep learning in vision can be attributed to: (a) models with high capacity; (b) increased computational power; and (c) availability of large-scale labeled data. Since 2012, there have been significant advances in representation capabilities of the models and computational capabilities of GPUs. But the size of the biggest dataset has surprisingly remained constant. What will happen if we increase the dataset size by 10x or 100x? This paper takes a step towards clearing the clouds of mystery surrounding the relationship between `enormous data' and visual deep learning. By exploiting the JFT-300M dataset which has more than 375M noisy labels for 300M images, we investigate how the performance of current vision tasks would change if this data was used for representation learning. Our paper delivers some surprising (and some expected) findings. First, we find that the performance on vision tasks increases logarithmically based on volume of training data size. Second, we show that representation learning (or pre-training) still holds a lot of promise. One can improve performance on many vision tasks by just training a better base model. Finally, as expected, we present new state-of-the-art results for different vision tasks including image classification, object detection, semantic segmentation and human pose estimation. Our sincere hope is that this inspires vision community to not undervalue the data and develop collective efforts in building larger datasets.

Image geolocalization, inferring the geographic location of an image, is a challenging computer vision problem with many potential applications. The recent state-of-the-art approach to this problem is a deep image classification approach in which the world is spatially divided into bins and a deep network is trained to predict the correct bin for a given image. We propose to combine this approach with the original Im2GPS approach in which a query image is matched against a database of geotagged images and the location is inferred from the retrieved set. We estimate the geographic location of a query image by applying kernel density estimation to the locations of its nearest neighbors in the reference database. Interestingly, we find that the best features for our retrieval task are derived from networks trained with classification loss even though we do not use a classification approach at test time. Training with classification loss outperforms several deep feature learning methods (e.g. Siamese networks with contrastive of triplet loss) more typical for retrieval applications. Our simple approach achieves state-of-the-art geolocalization accuracy while also requiring significantly less training data.

Image aberrations can cause severe degradation in image quality for consumer-level cameras, especially under the current tendency to reduce the complexity of lens designs in order to shrink the overall size of modules. In simplified optical designs, chromatic aberration can be one of the most significant causes for degraded image quality, and it can be quite difficult to remove in post-processing, since it results in strong blurs in at least some of the color channels. In this work, we revisit the pixel-wise similarity between different color channels of the image and accordingly propose a novel algorithm for correcting chromatic aberration based on this cross-channel correlation. In contrast to recent weak prior-based models, ours uses strong pixel-wise fitting and transfer, which lead to significant quality improvements for large chromatic aberrations. Experimental results on both synthetic and real world images captured by different optical systems demonstrate that the chromatic aberration can be significantly reduced using our approach.

An emerging problem in computer vision is the reconstruction of 3D shape and pose of an object from a single image. Hitherto, the problem has been addressed through the application of canonical deep learning methods to regress from the image directly to the 3D shape and pose labels. These approaches, however, are problematic from two perspectives. First, they are minimizing the error between 3D shapes and pose labels - with little thought about the nature of this "label error" when reprojecting the shape back onto the image. Second, they rely on the onerous and ill-posed task of hand labeling natural images with respect to 3D shape and pose. In this paper we define the new task of pose-aware shape reconstruction from a single image, and we advocate that cheaper 2D annotations of objects silhouettes in natural images can be utilized. We design architectures of pose-aware shape reconstruction which reproject the predicted shape back on to the image using the predicted pose. Our evaluation on several object categories demonstrates the superiority of our method for predicting pose-aware 3D shapes from natural images.

We introduce a novel method for representation learning that uses an artificial supervision signal based on counting visual primitives. This supervision signal is obtained from an equivariance relation, which does not require any manual annotation. We relate transformations of images to transformations of the representations. More specifically, we look for the representation that satisfies such relation rather than the transformations that match a given representation. In this paper, we use two image transformations in the context of counting: scaling and tiling. The first transformation exploits the fact that the number of visual primitives should be invariant to scale. The second transformation allows us to equate the total number of visual primitives in each tile to that in the whole image. These two transformations are combined in one constraint and used to train a neural network with a contrastive loss. The proposed task produces representations that perform on par or exceed the state of the art in transfer learning benchmarks.

The interactive image segmentation model allows users to iteratively add new inputs for refinement until a satisfactory result is finally obtained. Therefore, an ideal interactive segmentation model should learn to capture the user's intention with minimal interaction. However, existing models fail to fully utilize the valuable user input information in the segmentation refinement process and thus offer an unsatisfactory user experience. In order to fully exploit the user-provided information, we propose a new deep framework, called Regional Interactive Segmentation Network (RIS-Net), to expand the field-of-view of the given inputs to capture the local regional information surrounding them for local refinement. Additionally, RIS-Net adopts multiscale global contextual information to augment each local region for improving feature representation. We also introduce click discount factors to develop a novel optimization strategy for more effective end-to-end training. Comprehensive evaluations on four challenging datasets well demonstrate the superiority of the proposed RIS-Net over other state-of-the-art approaches.

Region-based correspondence (RBC) is a highly relevant and non-trivial computer vision problem. Given two 3D shapes, RBC seeks segments/regions on these shapes that can be reliably put in correspondence. The problem thus consists both in finding the regions and determining the correspondences between them. This problem statement is similar to that of "biclustering", implying that RBC can be cast as a biclustering problem. Here, we exploit this implication by tackling RBC via a novel biclustering approach, called S4B (spatially smooth spike and slab biclustering), which: (i) casts the problem in a probabilistic low-rank matrix factorization perspective; (ii) uses a spike and slab prior to induce sparsity; (iii) is enriched with a spatial smoothness prior, based on geodesic distances, encouraging nearby vertices to belong to the same bicluster. This type of spatial prior cannot be used in classical biclustering techniques. We test the proposed approach on the FAUST dataset, outperforming both state-of-the-art RBC techniques and classical biclustering methods.

Recovering 3D scene geometry from underwater images involves the Refractive Structure-from-Motion (RSfM) problem, where the image distortions caused by light refraction at the interface between different propagation media invalidates the single view point assumption. Direct use of the pinhole camera model in RSfM leads to inaccurate camera pose estimation and consequently drift. RSfM methods have been thoroughly studied for the case of a thick glass interface that assumes two refractive interfaces between the camera and the viewed scene. On the other hand, when the camera lens is in direct contact with the water, there is only one refractive interface. By explicitly considering a refractive interface, we develop a succinct derivation of the refractive fundamental matrix in the form of the generalised epipolar constraint for an axial camera. We use the refractive fundamental matrix to refine initial pose estimates obtained by assuming the pinhole model. This strategy allows us to robustly estimate underwater camera poses, where other methods suffer from poor noise-sensitivity. We also formulate a new four view constraint enforcing camera pose consistency along a video which leads us to a novel RSfM framework. For validation we use synthetic data to show the numerical properties of our method and we provide results on real data to demonstrate performance within laboratory settings and for applications in endoscopy.

We propose the use of a light-weight setup consisting of a collocated camera and light source --- commonly found on mobile devices --- to reconstruct surface normals and spatially-varying BRDFs of near-planar material samples. A collocated setup provides only a 1-D "univariate" sampling of the 4-D BRDF. We show that a univariate sampling is sufficient to estimate parameters of commonly used analytical BRDF models. Subsequently, we use a dictionary-based reflectance prior to derive a robust technique for per-pixel normal and BRDF estimation. We demonstrate real-world shape and capture, and its application to material editing and classification, using real data acquired using a mobile phone.

Referring expression is a kind of language expression that used for referring to particular objects.To make the expression without ambiguation, people often use attributes to describe the particular object. In this paper, we explore the role of attributes by incorporating them into both referring expression generation and comprehension. We first train an attribute learning model from visual objects and their paired descriptions. Then in the generation task, we take the learned attributes as the input into the generation model, thus the expressions are generated driven by both attributes and the previous words. For comprehension, we embed the learned attributes with visual features and semantics into the common space model, then the target object is retrieved based on its ranking distance in the common space. Experimental results on the three standard datasets, RefCOCO, RefCOCO+, and RefCOCOg show significant improvements over the baseline model, demonstrating that our methods are effective for both tasks.

Convolutional Neural Network (CNN) has led to significant progress in face recognition. Currently most CNN-based face recognition methods follow a two-step pipeline, i.e. a detected face is first aligned to a canonical one pre-defined by a mean face shape, and then it is fed into a CNN to extract features for recognition. The alignment step transforms all faces to the same shape, which can cause loss of geometrical information which is helpful in distinguishing different subjects. Moreover, it is hard to define a single optimal shape for the following recognition, since faces have large diversity in facial features, e.g. poses, illumination, etc. To be free from the above problems with an independent alignment step, we introduce a Recursive Spatial Transformer (ReST) module into CNN, allowing face alignment to be jointly learned with face recognition in an end-to-end fashion. The designed ReST has an intrinsic recursive structure and is capable of progressively aligning faces to a canonical one, even those with large variations. To model non-rigid transformation, multiple ReST modules are organized in a hierarchical structure to account for different parts of faces. Overall, the proposed ReST can handle large face variations and non-rigid transformation, and is end-to-end learnable and adaptive to input, making it an effective alignment-free face recognition solution. Extensive experiments are performed on LFW and YTF datasets, and the proposed ReST outperforms those two-step methods, demonstrating its effectiveness.