ICCV 2019 Papers

This paper addresses the generation of referring expressions that not only refer to objects correctly but also let humans find them quickly. As a target becomes relatively less salient, identifying referred objects itself becomes more difficult. However, the existing studies regarded all sentences that refer to objects correctly as equally good, ignoring whether they are easily understood by humans. If the target is not salient, humans utilize relationships with the salient contexts around it to help listeners to comprehend it better. To derive this information from human annotations, our model is designed to extract information from the target and from the environment. Moreover, we regard that sentences that are easily understood are those that are comprehended correctly and quickly by humans. We optimized this by using the time required to locate the referred objects by humans and their accuracies. To evaluate our system, we created a new referring expression dataset whose images were acquired from Grand Theft Auto V (GTA V), limiting targets to persons. Experimental results show the effectiveness of our approach. Our code and dataset are available at https://github.com/mikittt/easy-to-understand-REG.

Recently there has been significant progress in image captioning with the help of deep learning. However, captions generated by current state-of-the-art models are still far from satisfactory, despite high scores in terms of conventional metrics such as BLEU and CIDEr. Human-written captions are diverse, informative and precise, but machine-generated captions seem to be simple, vague and dull. In this paper, aimed at improving diversity and descriptiveness characteristics of generated image captions, we propose a model utilizing visual paraphrases (different sentences describing the same image) in captioning datasets. We explore different strategies to select useful visual paraphrase pairs for training by designing a variety of scoring functions. Our model consists of two decoding stages, where a preliminary caption is generated in the first stage and then paraphrased into a more diverse and descriptive caption in the second stage. Extensive experiments are conducted on the benchmark MS COCO dataset, with automatic evaluation and human evaluation results verifying the effectiveness of our model.

Understanding where people are looking is an informative social cue. In this work, we present Gaze360, a large-scale remote gaze-tracking dataset and method for robust 3D gaze estimation in unconstrained images. Our dataset consists of 238 subjects in indoor and outdoor environments with labelled 3D gaze across a wide range of head poses and distances. It is the largest publicly available dataset of its kind by both subject and variety, made possible by a simple and efficient collection method. Our proposed 3D gaze model extends existing models to include temporal information and to directly output an estimate of gaze uncertainty. We demonstrate the benefits of our model via an ablation study, and show its generalization performance via a cross-dataset evaluation against other recent gaze benchmark datasets. We furthermore propose a simple self-supervised approach to improve cross-dataset domain adaptation. Finally, we demonstrate an application of our model for estimating customer attention in a supermarket setting. Our dataset and models will be made available at http://gaze360.csail.mit.edu.

The use of object detection algorithms is becoming increasingly important in autonomous vehicles, and object detection at high accuracy and a fast inference speed is essential for safe autonomous driving. A false positive (FP) from a false localization during autonomous driving can lead to fatal accidents and hinder safe and efficient driving. Therefore, a detection algorithm that can cope with mislocalizations is required in autonomous driving applications. This paper proposes a method for improving the detection accuracy while supporting a real-time operation by modeling the bounding box (bbox) of YOLOv3, which is the most representative of one-stage detectors, with a Gaussian parameter and redesigning the loss function. In addition, this paper proposes a method for predicting the localization uncertainty that indicates the reliability of bbox. By using the predicted localization uncertainty during the detection process, the proposed schemes can significantly reduce the FP and increase the true positive (TP), thereby improving the accuracy. Compared to a conventional YOLOv3, the proposed algorithm, Gaussian YOLOv3, improves the mean average precision (mAP) by 3.09 and 3.5 on the KITTI and Berkeley deep drive (BDD) datasets, respectively. Nevertheless, the proposed algorithm is capable of real-time detection at faster than 42 frames per second (fps) and shows a higher accuracy than previous approaches with a similar fps. Therefore, the proposed algorithm is the most suitable for autonomous driving applications.

Real-world object classes appear in imbalanced ratios. This poses a significant challenge for classifiers which get biased towards frequent classes. We hypothesize that improving the generalization capability of a classifier should improve learning on imbalanced datasets. Here, we introduce the first hybrid loss function that jointly performs classification and clustering in a single formulation. Our approach is based on an `affinity measure' in Euclidean space that leads to the following benefits: (1) direct enforcement of maximum margin constraints on classification boundaries, (2) a tractable way to ensure uniformly spaced and equidistant cluster centers, (3) flexibility to learn multiple class prototypes to support diversity and discriminability in feature space. Our extensive experiments demonstrate the significant performance improvements on visual classification and verification tasks on multiple imbalanced datasets. The proposed loss can easily be plugged in any deep architecture as a differentiable block and demonstrates robustness against different levels of data imbalance and corrupted labels.

Current state-of-the-art methods for image segmentation form a dense image representation where the color, shape and texture information are all processed together inside a deep CNN. This however may not be ideal as they contain very different type of information relevant for recognition. Here, we propose a new two-stream CNN architecture for semantic segmentation that explicitly wires shape information as a separate processing branch, i.e. shape stream, that processes information in parallel to the classical stream. Key to this architecture is a new type of gates that connect the intermediate layers of the two streams. Specifically, we use the higher-level activations in the classical stream to gate the lower-level activations in the shape stream, effectively removing noise and helping the shape stream to only focus on processing the relevant boundary-related information. This enables us to use a very shallow architecture for the shape stream that operates on the image-level resolution. Our experiments show that this leads to a highly effective architecture that produces sharper predictions around object boundaries and significantly boosts performance on thinner and smaller objects. Our method achieves state-of-the-art performance on the Cityscapes benchmark, in terms of both mask (mIoU) and boundary (F-score) quality, improving by 2% and 4% over strong baselines.

While Physics-Based Simulation (PBS) can accurately drape a 3D garment on a 3D body, it remains too costly for real-time applications, such as virtual try-on. By contrast, inference in a deep network, requiring a single forward pass, is much faster. Taking advantage of this, we propose a novel architecture to fit a 3D garment template to a 3D body. Specifically, we build upon the recent progress in 3D point cloud processing with deep networks to extract garment features at varying levels of detail, including point-wise, patch-wise and global features. We fuse these features with those extracted in parallel from the 3D body, so as to model the cloth-body interactions. The resulting two-stream architecture, which we call as GarNet, is trained using a loss function inspired by physics-based modeling, and delivers visually plausible garment shapes whose 3D points are, on average, less than 1 cm away from those of a PBS method, while running 100 times faster. Moreover, the proposed method can model various garment types with different cutting patterns when parameters of those patterns are given as input to the network.

SLAM technology has recently seen many successes and attracted the attention of high-technological companies. However, how to unify the interface of existing or emerging algorithms, and effectively perform benchmark about the speed, robustness and portability are still problems. In this paper, we propose a novel SLAM platform named GSLAM, which not only provides evaluation functionality, but also supplies useful toolkit for researchers to quickly develop their SLAM systems. Our core contribution is an universal, cross-platform and full open-source SLAM interface for both research and commercial usage, which is aimed to handle interactions with input dataset, SLAM implementation, visualization and applications in an unified framework. Through this platform, users can implement their own functions for better performance with plugin form and further boost the application to practical usage of the SLAM.

We present a joint 3D pose and focal length estimation approach for object categories in the wild. In contrast to previous methods that predict 3D poses independently of the focal length or assume a constant focal length, we explicitly estimate and integrate the focal length into the 3D pose estimation. For this purpose, we combine deep learning techniques and geometric algorithms in a two-stage approach: First, we estimate an initial focal length and establish 2D-3D correspondences from a single RGB image using a deep network. Second, we recover 3D poses and refine the focal length by minimizing the reprojection error of the predicted correspondences. In this way, we exploit the geometric prior given by the focal length for 3D pose estimation. This results in two advantages: First, we achieve significantly improved 3D translation and 3D pose accuracy compared to existing methods. Second, our approach finds a geometric consensus between the individual projection parameters, which is required for precise 2D-3D alignment. We evaluate our proposed approach on three challenging real-world datasets (Pix3D, Comp, and Stanford) with different object categories and significantly outperform the state-of-the-art by up to 20% absolute in multiple different metrics.

One-class learning is the classic problem of fitting a model to data for which annotations are available only for a single class. In this paper, we propose a novel objective for one-class learning. Our key idea is to use a pair of orthonormal frames -- as subspaces -- to "sandwich" the labeled data via optimizing for two objectives jointly: i) minimize the distance between the origins of the two subspaces, and ii) to maximize the margin between the hyperplanes and the data, either subspace demanding the data to be in its positive and negative orthant respectively. Our proposed objective however leads to a non-convex optimization problem, to which we resort to Riemannian optimization schemes and derive an efficient conjugate gradient scheme on the Stiefel manifold. To study the effectiveness of our scheme, we propose a new dataset Dash-Cam-Pose, consisting of clips with skeleton poses of humans seated in a car, the task being to classify the clips as normal or abnormal; the latter is when any human pose is out-of-position with regard to say an airbag deployment. Our experiments on the proposed Dash-Cam-Pose dataset, as well as several other standard anomaly/novelty detection benchmarks demonstrate the benefits of our scheme, achieving state-of-the-art one-class accuracy.

At the core of most three-dimensional alignment and tracking tasks resides the critical problem of point correspondence. In this context, the design of descriptors that efficiently and uniquely identifies keypoints, to be matched, is of central importance. Numerous descriptors have been developed for dealing with affine/perspective warps, but few can also handle non-rigid deformations. In this paper, we introduce a novel binary RGB-D descriptor invariant to isometric deformations. Our method uses geodesic isocurves on smooth textured manifolds. It combines appearance and geometric information from RGB-D images to tackle non-rigid transformations. We used our descriptor to track multiple textured depth maps and demonstrate that it produces reliable feature descriptors even in the presence of strong non-rigid deformations and depth noise. The experiments show that our descriptor outperforms different state-of-the-art descriptors in both precision-recall and recognition rate metrics. We also provide to the community a new dataset composed of annotated RGB-D images of different objects (shirts, cloths, paintings, bags), subjected to strong non-rigid deformations, to evaluate point correspondence algorithms.

We introduce a framework that uses Generative Adversarial Networks (GANs) to study cognitive properties like memorability. These attributes are of interest because we do not have a concrete visual definition of what they entail. What does it look like for a dog to be more memorable? GANs allow us to generate a manifold of natural-looking images with fine-grained differences in their visual attributes. By navigating this manifold in directions that increase memorability, we can visualize what it looks like for a particular generated image to become more memorable. The resulting "visual definitions" surface image properties (like "object size") that may underlie memorability. Through behavioral experiments, we verify that our method indeed discovers image manipulations that causally affect human memory performance. We further demonstrate that the same framework can be used to analyze image aesthetics and emotional valence. ganalyze.csail.mit.edu.

Despite the remarkable success of generative adversarial networks, their performance seems less impressive for diverse training sets, requiring learning of discontinuous mapping functions. Though multi-mode prior or multi-generator models have been proposed to alleviate this problem, such approaches may fail depending on the empirically chosen initial mode components. In contrast to such bottom-up approaches, we present GAN-Tree, which follows a hierarchical divisive strategy to address such discontinuous multi-modal data. Devoid of any assumption on the number of modes, GAN-Tree utilizes a novel mode-splitting algorithm to effectively split the parent mode to semantically cohesive children modes, facilitating unsupervised clustering. Further, it also enables incremental addition of new data modes to an already trained GAN-Tree, by updating only a single branch of the tree structure. As compared to prior approaches, the proposed framework offers a higher degree of flexibility in choosing a large variety of mutually exclusive and exhaustive tree nodes called GAN-Set. Extensive experiments on synthetic and natural image datasets including ImageNet demonstrate the superiority of GAN-Tree against the prior state-of-the-art.

A Generative Adversarial Network (GAN) with generator G trained to model the prior of images has been shown to perform better than sparsity-based regularizers in ill-posed inverse problems. Here, we propose a new method of deploying a GAN-based prior to solve linear inverse problems using projected gradient descent (PGD). Our method learns a network-based projector for use in the PGD algorithm, eliminating expensive computation of the Jacobian of G. Experiments show that our approach provides a speed-up of 60-80x over earlier GAN-based recovery methods along with better accuracy in compressed sensing. Our main theoretical result is that if the measurement matrix is moderately conditioned on the manifold range(G) and the projector is d-approximate, then the algorithm is guaranteed to reach O(d) reconstruction error in O(log(1/d)) steps in the low noise regime. Additionally, we propose a fast method to design such measurement matrices for a given G. Extensive experiments demonstrate the efficacy of this method by requiring 5-10x fewer measurements than random Gaussian measurement matrices for comparable recovery performance. Because the learning of the GAN and projector is decoupled from the measurement operator, our GAN-based projector and recovery algorithm are applicable without retraining to all linear inverse problems in which the measurement operator is moderately conditioned for range(G), as confirmed by experiments on compressed sensing, super-resolution, and inpainting.

Recent adversarial learning research has achieved very impressive progress for modelling cross-domain data shifts in appearance space but its counterpart in modelling cross-domain shifts in geometry space lags far behind. This paper presents an innovative Geometry-Aware Domain Adaptation Network (GA-DAN) that is capable of modelling cross-domain shifts concurrently in both geometry space and appearance space and realistically converting images across domains with very different characteristics. In the proposed GA-DAN, a novel multi-modal spatial learning structure is designed which can convert a source-domain image into multiple images of different spatial views as in the target domain. A new disentangled cycle-consistency loss is introduced which balances the cycle consistency and greatly improves the concurrent learning in both appearance and geometry spaces. The proposed GA-DAN has been evaluated for the classic scene text detection and recognition tasks, and experiments show that the domain-adapted images achieve superior scene text detection and recognition performance while applied to network training.

In this paper we present an end-to-end framework for grounding of phrases in images. In contrast to previous works, our model, which we call GraphGround, uses graphs to formulate more complex, non-sequential dependencies among proposal image regions and phrases. We capture intra-modal dependencies using a separate graph neural network for each modality (visual and lingual), and then use conditional message-passing in another graph neural network to fuse their outputs and capture cross-modal relationships. This final representation results in grounding decisions. The framework supports many-to-many matching and is able to ground single phrase to multiple image regions and vice versa. We validate our design choices through a series of ablation studies and illustrate state-of-the-art performance on Flickr30k and ReferIt Game benchmark datasets.

We propose a new image denoising algorithm, dubbed as Fully Convolutional Adaptive Image DEnoiser (FC-AIDE), that can learn from an offline supervised training set with a fully convolutional neural network as well as adaptively fine-tune the supervised model for each given noisy image. We significantly extend the framework of the recently proposed Neural AIDE, which formulates the denoiser to be context-based pixelwise mappings and utilizes the unbiased estimator of MSE for such denoisers. The two main contributions we make are; 1) implementing a novel fully convolutional architecture that boosts the base supervised model, and 2) introducing regularization methods for the adaptive fine-tuning such that a stronger and more robust adaptivity can be attained. As a result, FC-AIDE is shown to possess many desirable features; it outperforms the recent CNN-based state-of-the-art denoisers on all of the benchmark datasets we tested, and gets particularly strong for various challenging scenarios, e.g., with mismatched image/noise characteristics or with scarce supervised training data. The source code our algorithm is available at https://github.com/csm9493/FC-AIDE-Keras https://github.com/csm9493/FC-AIDE-Keras .

Extracting geometric features from 3D scans or point clouds is the first step in applications such as registration, reconstruction, and tracking. State-of-the-art methods require computing low-level features as input or extracting patch-based features with limited receptive field. In this work, we present fully-convolutional geometric features, computed in a single pass by a 3D fully-convolutional network. We also present new metric learning losses that dramatically improve performance. Fully-convolutional geometric features are compact, capture broad spatial context, and scale to large scenes. We experimentally validate our approach on both indoor and outdoor datasets. Fully-convolutional geometric features achieve state-of-the-art accuracy without requiring prepossessing, are compact (32 dimensions), and are 290 times faster than the most accurate prior method.

Visual counting, a task that predicts the number of objects from an image/video, is an open-set problem by nature, i.e., the number of population can vary in [0,+[?]) in theory. However, the collected images and labeled count values are limited in reality, which means only a small closed set is observed. Existing methods typically model this task in a regression manner, while they are likely to suffer from an unseen scene with counts out of the scope of the closed set. In fact, counting is decomposable. A dense region can always be divided until the count values of sub-regions are within the previously observed closed set. Inspired by this idea, we propose a simple but effective approach, Spatial Divide-and-Conquer Network (S-DCNet). S-DCNet learns to classify closed-set counts and can generalize to open-set counts via S-DC. S-DCNet is also efficient. To avoid repeatedly computing sub-region convolutional features, S-DC is executed on the feature map instead of on the input image. S-DCNet achieves the state-of-the-art performance on three crowd counting datasets (ShanghaiTech, UCF_CC_50 and UCF-QNRF), a vehicle counting dataset (TRANCOS) and a plant counting dataset (MTC). Compared to the previous best methods, S-DCNet brings a 20.2% relative improvement on the ShanghaiTechPart B, 20.9% on the UCF-QNRF, 22.5% on the TRANCOS and 15.1% on the MTC. Code has been made available at: https://github.com/xhp-hust-2018-2011/S-DCNet.