ICCV 2017 Papers

We consider retrieving a specific temporal segment, or moment, from a video given a natural language text description. Methods designed to retrieve whole video clips with natural language determine what occurs in a video but not when. To address this issue, we propose the Moment Context Network (MCN) which effectively localizes natural language queries in videos by integrating local and global video features over time. A key obstacle to training our MCN model is that current video datasets do not include pairs of localized video segments and referring expressions, or text descriptions which uniquely identify a corresponding moment. Therefore, we collect the Distinct Describable Moments (DiDeMo) dataset which consists of over 10,000 unedited, personal videos in diverse visual settings with pairs of localized video segments and referring expressions. We demonstrate that MCN outperforms several baseline methods and believe that our initial results together with release of DiDeMo will inspire further research on localizing video moments with natural language.

Local-to global point cloud registration is a challenging task due to the substantial differences between these two types of data, and the different techniques used to acquire them. Global clouds cover large-scale environments and are usually acquired aerially, e.g., 3D modeling of a city using Airborne Laser Scanning (ALS). In contrast, local clouds are often acquired from ground level and at a much smaller range, for example, using Terrestrial Laser Scanning (TLS). The differences are often manifested in point density distribution, occlusions nature, and measurement noise. As a result of these differences, existing point cloud registration approaches, such as keypoint-based registration, tend to fail. We improve upon a different approach, recently proposed, based on converting the global cloud into a viewpoint-based cloud dictionary. We propose a local-to-global registration method where we replace the dictionary clouds with viewpoint descriptors, consisting of panoramic range-images. We then use an efficient dictionary search in the Discrete Fourier Transform (DFT) domain, using phase correlation, to rapidly find plausible transformations from the local to the global reference frame. We demonstrate our method's significant advantages over the previous cloud dictionary approach, in terms of computational efficiency and memory requirements. In addition, We show its superior registration performance in comparison to a state-of-the-art, keypoint-based method (FPFH). For the evaluation, we use a challenging dataset of TLS local clouds and an ALS large-scale global cloud, in an urban environment.

In this paper we present a differential approach to photo-polarimetric shape estimation. We propose several alternative differential constraints based on polarisation and photometric shading information and show how to express them in a unified partial differential system. Our method uses the image ratios technique to combine shading and polarisation information in order to directly reconstruct surface height, without first computing surface normal vectors. Moreover, we are able to remove the non-linearities so that the problem reduces to solving a linear differential problem. We also introduce a new method for estimating a polarisation image from multichannel data and, finally, we show it is possible to estimate the illumination directions in a two source setup, extending the method into an uncalibrated scenario. From a numerical point of view, we use a least-squares formulation of the discrete version of the problem. To the best of our knowledge, this is the first work to consider a unified differential approach to solve photo-polarimetric shape estimation directly for height. Numerical results on synthetic and real-world data confirm the effectiveness of our proposed method.

Existing video event classification approaches suffer from limited human-labeled semantic annotations. Weak semantic annotations can be harvested from Web-knowledge without involving any human interaction. However such weak annotations are noisy, thus can not be effectively utilized without distinguishing its reliability. In this paper, we propose a novel approach to automatically maximize the utility of weak semantic annotations (formalized as the semantic relevance of video shots to the target event) to facilitate video event classification. A novel attention model is designed to determine the attention scores of video shots, where the weak semantic relevance is considered as attentional guidance. Specifically, our model jointly optimizes two objectives at different levels. The first one is the classification loss corresponding to video-level groundtruth labels, and the second is the shot-level relevance loss corresponding to weak semantic relevance. We use a long short-term memory (LSTM) layer to capture the temporal information carried by the shots of a video. In each timestep, the LSTM employs the attention model to weight the current shot under the guidance of its weak semantic relevance to the event of interest. Thus, we can automatically exploit weak semantic relevance to assist video event classification. Extensive experiments have been conducted on three complex large-scale video event datasets i.e., MEDTest14, ActivityNet and FCVID. Our approach achieves the state-of-the-art classification performance on all three datasets. The significant performance improvement upon the conventional attention model also demonstrates the effectiveness of our model.

Unsupervised learning with generative adversarial networks (GANs) has proven hugely successful. Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function. However, we found that this loss function may lead to the vanishing gradients problem during the learning process. To overcome such a problem, we propose in this paper the Least Squares Generative Adversarial Networks (LSGANs) which adopt the least squares loss function for the discriminator. We show that minimizing the objective function of LSGAN yields minimizing the Pearson Chi^2 divergence. There are two benefits of LSGANs over regular GANs. First, LSGANs are able to generate higher quality images than regular GANs. Second, LSGANs perform more stable during the learning process. We evaluate LSGANs on LSUN and CIFAR-10 datasets and the experimental results show that the images generated by LSGANs are of better quality than the ones generated by regular GANs. We also conduct two comparison experiments between LSGANs and regular GANs to illustrate the stability of LSGANs.

Image understanding enables better reconstruction of the physical world from images and videos. Existing methods focus largely on geometry and visual appearance of the reconstructed scene. In this paper, we extend the frontier in image understanding and present a new technique to recover the material properties of cloth from a video.Previous cloth material recovery methods often require markers or complex experimental set-up to acquire physical properties, or are limited to certain types of images/videos. Our approach takes advantages of the appearance changes of the moving cloth to infer its physical properties. To extract information about the cloth, our method characterizes both the motion space and the visual appearance of the cloth geometry. We apply the Convolutional Neural Network (CNN) and the Long Short Term Memory (LSTM) neural network to material recovery of cloth properties from videos. We also exploit simulated data to help statistical learning of mapping between the visual appearance and motion dynamics of the cloth. The effectiveness of our method is demonstrated via validation using simulated datasets and real-life recorded videos.

We present a machine learning algorithm that takes as input a 2D RGB image and synthesizes a 4D RGBD light field (color and depth of the scene in each ray direction). For training, we introduce the largest public light field dataset, consisting of over 3300 plenoptic camera light fields of scenes containing flowers and plants. Our synthesis pipeline consists of a convolutional neural network (CNN) that estimates scene geometry, a stage that renders a Lambertian light field using that geometry, and a second CNN that predicts occluded rays and non-Lambertian effects. Our algorithm builds on recent view synthesis methods, but is unique in predicting RGBD for each light field ray and improving unsupervised single image depth estimation by enforcing consistency of ray depths that should intersect the same scene point.

We present an algorithm to directly restore a clear high-resolution image from a blurry low-resolution input. This problem is highly ill-posed and the basic assumptions for existing super-resolution methods (requiring clear input) and deblurring methods (requiring high-resolution input) no longer hold. We focus on face and text images and adopt a generative adversarial network (GAN) to learn a category-specific prior to solve this problem. However, the basic GAN formulation does not generate realistic high-resolution images. In this work, we introduce novel training losses that help recover fine details. We also present a multi-class GAN that can process multi-class image restoration tasks, i.e., face and text images, using a single generator network. Extensive experiments demonstrate that our method performs favorably against the state-of-the-art methods on both synthetic and real-world images at a lower computational cost.

Natural language questions are inherently compositional, and many are most easily answered by reasoning about their decomposition into modular sub-problems. For example, to answer "is there an equal number of balls and boxes?" we can look for balls, look for boxes, count them, and compare the results. The recently proposed Neural Module Network (NMN) architecture implements this approach to question answering by parsing questions into linguistic substructures and assembling question-specific deep networks from smaller modules that each solve one subtask. However, existing NMN implementations rely on brittle off-the-shelf parsers, and are restricted to the module configurations proposed by these parsers rather than learning them from data. In this paper, we propose End-to-End Module Networks (N2NMNs), which learn to reason by directly predicting instance-specific network layouts without the aid of a parser. Our model learns to generate network structures (by imitating expert demonstrations) while simultaneously learning network parameters (using the downstream task loss). Experimental results on the new CLEVR dataset targeted at compositional question answering show that N2NMNs achieve an error reduction of nearly 50% relative to state-of-the-art attentional approaches, while discovering interpretable network architectures specialized for each question.

This work addresses the task of non-blind image deconvolution. Motivated to keep up with the constant increase in image size, with megapixel images becoming the norm, we aim at pushing the limits of efficient FFT-based techniques. Based on an analysis of traditional and more recent learning-based methods, we generalize existing discriminative approaches by using more powerful regularization, based on convolutional neural networks. Additionally, we propose a simple, yet effective, boundary adjustment method that alleviates the problematic circular convolution assumption, which is necessary for FFT-based deconvolution. We evaluate our approach on two common non-blind deconvolution benchmarks and achieve state-of-the-art results even when including methods which are computationally considerably more expensive.

Most recent approaches to monocular 3D human pose estimation rely on Deep Learning. They typically involve regressing from an image to either 3D joint coordinates directly or 2D joint locations from which 3D coordinates are inferred. Both approaches have their strengths and weaknesses and we therefore propose a novel architecture designed to deliver the best of both worlds by performing both simultaneously and fusing the information along the way. At the heart of our framework is a trainable fusion scheme that learns how to fuse the information optimally instead of being hand-designed. This yields significant improvements upon the state-of-the-art on standard 3D human pose estimation benchmarks.

Low-cost consumer depth cameras and deep learning have enabled reasonable 3D hand pose estimation from single depth images. In this paper, we present an approach that estimates 3D hand pose from regular RGB images. This task has far more ambiguities due to the missing depth information. To this end, we propose a deep network that learns a network-implicit 3D articulation prior. Together with detected keypoints in the images, this network yields good estimates of the 3D pose. We introduce a large scale 3D hand pose dataset based on synthetic hand models for training the involved networks. Experiments on a variety of test sets, including one on sign language recognition, demonstrate the feasibility of 3D hand pose estimation on single color images.

The ability to ask questions is a powerful tool to gather information in order to learn about the world and resolve ambiguities. In this paper, we explore a novel problem of generating discriminative questions to help disambiguate visual instances. Our work can be seen as a complement and new extension to the rich research studies on image captioning and question answering. We introduce the first large-scale dataset with over 10,000 carefully annotated images-question tuples to facilitate benchmarking. In particular, each tuple consists of a pair of images and 4.6 discriminative questions (as positive samples) and 5.9 non-discriminative questions (as negative samples) on average. In addition, we present an effective method for visual discriminative question generation. The method can be trained in a weakly supervised manner without discriminative images-question tuples but just existing visual question answering datasets. Promising results are shown against representative baselines through quantitative evaluations and user studies.

What defines a visual style? Fashion styles emerge organically from how people assemble outfits of clothing, making them difficult to pin down with a computational model. Low-level visual similarity can be too specific to detect stylistically similar images, while manually crafted style categories can be too abstract to capture subtle style differences. We propose an unsupervised approach to learn a style-coherent representation. Our method leverages probabilistic polylingual topic models based on visual attributes to discover a set of latent style factors. Given a collection of unlabeled fashion images, our approach mines for the latent styles, then summarizes outfits by how they mix those styles. Our approach can organize galleries of outfits by style without requiring any style labels. Experiments on over 100K images demonstrate its promise for retrieving, mixing, and summarizing fashion images by their style.

Many prediction tasks contain uncertainty. In some cases, uncertainty is inherent in the task itself. In future prediction, for example, many distinct outcomes are equally valid. In other cases, uncertainty arises from the way data is labeled. For example, in object detection, many objects of interest often go unlabeled, and in human pose estimation, occluded joints are often labeled with ambiguous values. In this work we focus on a principled approach for handling such scenarios. In particular, we propose a framework for reformulating existing single-prediction models as multiple hypothesis prediction (MHP) models and an associated meta loss and optimization procedure to train them. To demonstrate our approach, we consider four diverse applications: human pose estimation, future prediction, image classification and segmentation. We find that MHP models outperform their single-hypothesis counterparts in all cases, and that MHP models simultaneously expose valuable insights into the variability of predictions.

We propose an active 3D mapping method for depth sensors, which allow individual control of depth-measuring rays, such as the newly emerging Solid State Lidars. The method simultaneously (i) learns to reconstruct a dense 3D voxel-map from sparse depth measurements, and (ii) optimizes the reactive control of depth-measuring rays. To make the first step towards the online control optimization, we propose a fast greedy algorithm, which needs to update its cost function in only a small fraction of possible rays. The approximation ratio of the greedy algorithm is derived. Experimental evaluation on the subset of the Kitti dataset demonstrates significant improvement in the 3D map accuracy when learning-to-reconstruct from sparse measurements is coupled with the optimization where-to-measure.

A major challenge in matching between vision and language is that they typically have completely different features and representations. In this work, we introduce a novel bridge between the modality-specific representations by creating a co-embedding space based on a recurrent residual fusion (RRF) block. Specifically, RRF adapts the recurrent mechanism to residual learning, so that it can recursively improve feature embeddings while retaining the shared parameters. Then, a fusion module is used to integrate the intermediate recurrent outputs and generates a more powerful representation. In the matching network, RRF acts as a feature enhancement component to gather visual and textual representations into a more discriminative embedding space where it allows to narrow the cross-modal gap between vision and language. Moreover, we employ a bi-rank loss function to enforce separability of the two modalities in the embedding space. In the experiments, we evaluate the proposed RRF-Net using two multi-modal datasets where it achieves state-of-the-art results.

Real-world image recognition systems need to recognize tens of thousands of classes that constitute a plethora of visual concepts. The traditional approach of annotating thousands of images per class for training is infeasible in such a scenario, prompting the use of webly supervised data. This paper explores the training of image-recognition systems on large numbers of images and associated user comments. In particular, we develop visual n-gram models that can predict arbitrary phrases that are relevant to the content of an image. Our visual n-gram models are feed-forward convolutional networks trained using new loss functions that are inspired by n-gram models commonly used in language modeling. We demonstrate the merits of our models in phrase prediction, phrase-based image retrieval, relating images and captions, and zero-shot transfer.

This paper presents a novel method for quantitative and objective diagnoses of Autism Spectrum Disorder (ASD) using eye tracking and deep neural networks. ASD is prevalent, with 1.5% of people in the US. The lack of clinical resources for early diagnoses has been a long-lasting issue. This work differentiates itself with three unique features: first, the proposed approach is data-driven and free of assumptions, important for new discoveries in understanding ASD as well as other neurodevelopmental disorders. Second, we concentrate our analyses on the differences in eye movement patterns between healthy people and those with ASD. An image selection method based on Fisher scores allows feature learning with the most discriminative contents, leading to efficient and accurate diagnoses. Third, we leverage the recent advances in deep neural networks for both prediction and visualization. Experimental results show the superior performance of our method in terms of multiple evaluation metrics used in diagnostic tests.