ICCV 2019 Papers

Obstacles facing accurate identification and localization of diseases in chest X-ray images lie in the lack of high-quality images and annotations. In this paper, we propose a Contrast Induced Attention Network (CIA-Net), which exploits the highly structured property of chest X-ray images and localizes diseases via contrastive learning on the aligned positive and negative samples. To force the attention module to focus only on sites of abnormalities, we also introduce a learnable alignment module to adjust all the input images, which eliminates variations of scales, angles, and displacements of X-ray images generated under bad scan conditions. We show that the use of contrastive attention and alignment module allows the model to learn rich identification and localization information using only a small amount of location annotations, resulting in state-of-the-art performance in NIH chest X-ray dataset.

Essential matrix averaging, i.e., the task of recovering camera locations and orientations in calibrated, multiview settings, is a first step in global approaches to Euclidean structure from motion. A common approach to essential matrix averaging is to separately solve for camera orientations and subsequently for camera positions. This paper presents a novel approach that solves simultaneously for both camera orientations and positions. We offer a complete characterization of the algebraic conditions that enable a unique Euclidean reconstruction of n cameras from a collection of (^n_2) essential matrices. We next use these conditions to formulate essential matrix averaging as a constrained optimization problem, allowing us to recover a consistent set of essential matrices given a (possibly partial) set of measured essential matrices computed independently for pairs of images. We finally use the recovered essential matrices to determine the global positions and orientations of the n cameras. We test our method on common SfM datasets, demonstrating high accuracy while maintaining efficiency and robustness, compared to existing methods.

Depth sensors like LIDARs and Kinect use a fixed depth acquisition strategy that is independent of the scene of interest. Due to the low spatial and temporal resolution of these sensors, this strategy can undersample parts of the scene that are important (small or fast moving objects), or oversample areas that are not informative for the task at hand (a fixed planar wall). In this paper, we present an approach and system to dynamically and adaptively sample the depths of a scene using the principle of triangulation light curtains. The approach directly detects the presence or absence of objects at specified 3D lines. These 3D lines can be sampled sparsely, non-uniformly, or densely only at specified regions. The depth sampling can be varied in real-time, enabling quick object discovery or detailed exploration of areas of interest. These results are achieved using a novel prototype light curtain system that is based on a 2D rolling shutter camera with higher light efficiency, working range, and faster adaptation than previous work, making it useful broadly for autonomous navigation and exploration.

Facial landmark detection, or face alignment, is a fundamental task that has been extensively studied. In this paper, we investigate a new perspective of facial landmark detection and demonstrate it leads to further notable improvement. Given that any face images can be factored into space of style that captures lighting, texture and image environment, and a style-invariant structure space, our key idea is to leverage disentangled style and shape space of each individual to augment existing structures via style translation. With these augmented synthetic samples, our semi-supervised model surprisingly outperforms the fully-supervised one by a large margin. Extensive experiments verify the effectiveness of our idea with state-of-the-art results on WFLW, 300W, COFW, and AFLW datasets. Our proposed structure is general and could be assembled into any face alignment frameworks. The code is made publicly available at https://github.com/thesouthfrog/stylealign.

It is well known that deep neural networks (DNNs) are vulnerable to adversarial attacks, which are implemented by adding crafted perturbations onto benign examples. Min-max robust optimization based adversarial training can provide a notion of security against adversarial attacks. However, adversarial robustness requires a significantly larger capacity of the network than that for the natural training with only benign examples. This paper proposes a framework of concurrent adversarial training and weight pruning that enables model compression while still preserving the adversarial robustness and essentially tackles the dilemma of adversarial training. Furthermore, this work studies two hypotheses about weight pruning in the conventional setting and finds that weight pruning is essential for reducing the network model size in the adversarial setting; training a small model from scratch even with inherited initialization from the large model cannot achieve neither adversarial robustness nor high standard accuracy. Code is available at https://github.com/yeshaokai/Robustness-Aware-Pruning-ADMM.

For many computer vision applications such as image captioning, visual question answering, and person search, learning discriminative feature representations at both image and text level is an essential yet challenging problem. Its challenges originate from the large word variance in the text domain as well as the difficulty of accurately measuring the distance between the features of the two modalities. Most prior work focuses on the latter challenge, by introducing loss functions that help the network learn better feature representations but fail to account for the complexity of the textual input. With that in mind, we introduce TIMAM: a Text-Image Modality Adversarial Matching approach that learns modality-invariant feature representations using adversarial and cross-modal matching objectives. In addition, we demonstrate that BERT, a publicly-available language model that extracts word embeddings, can successfully be applied in the text-to-image matching domain. The proposed approach achieves state-of-the-art cross-modal matching performance on four widely-used publicly-available datasets resulting in absolute improvements ranging from 2% to 5% in terms of rank-1 accuracy.

The Deep neural networks (DNNs) have achieved great success on a variety of computer vision tasks, however, they are highly vulnerable to adversarial attacks. To address this problem, we propose to improve the local smoothness of the representation space, by integrating a margin-based triplet embedding regularization term into the classification objective, so that the obtained models learn to resist adversarial examples. The regularization term consists of two steps optimizations which find potential perturbations and punish them by a large margin in an iterative way. Experimental results on MNIST, CASIA-WebFace, VGGFace2 and MS-Celeb-1M reveal that our approach increases the robustness of the network against both feature and label adversarial attacks in simple object classification and deep face recognition.

Recognizing unseen attribute-object pairs never appearing in the training data is a challenging task, since an object often refers to a specific entity while an attribute is an abstract semantic description. Besides, attributes are highly correlated to objects, i.e., an attribute tends to describe different visual features of various objects. Existing methods mainly employ two classifiers to recognize attribute and object separately, or simply simulate the composition of attribute and object, which ignore the inherent discrepancy and correlation between them. In this paper, we propose a novel adversarial fine-grained composition learning model for unseen attribute-object pair recognition. Considering their inherent discrepancy, we leverage multi-scale feature integration to capture discriminative fine-grained features from a given image. Besides, we devise a quintuplet loss to depict more accurate correlations between attributes and objects. Adversarial learning is employed to model the discrepancy and correlations among attributes and objects. Extensive experiments on two challenging benchmarks indicate that our method consistently outperforms state-of-the-art competitors by a large margin.

Thanks to their remarkable generative capabilities, GANs have gained great popularity, and are used abundantly in state-of-the-art methods and applications. In a GAN based model, a discriminator is trained to learn the real data distribution. To date, it has been used only for training purposes, where it's utilized to train the generator to provide real-looking outputs. In this paper we propose a novel method that makes an explicit use of the discriminator in test-time, in a feedback manner in order to improve the generator results. To the best of our knowledge it is the first time a discriminator is involved in test-time. We claim that the discriminator holds significant information on the real data distribution, that could be useful for test-time as well, a potential that has not been explored before. The approach we propose does not alter the conventional training stage. At test-time, however, it transfers the output from the generator into the discriminator, and uses feedback modules (convolutional blocks) to translate the features of the discriminator layers into corrections to the features of the generator layers, which are used eventually to get a better generator result. Our method can contribute to both conditional and unconditional GANs. As demonstrated by our experiments, it can improve the results of state-of-the-art networks for super-resolution, and image generation.

With the remarkable success of deep learning, Deep Neural Networks (DNNs) have been applied as dominant tools to various machine learning domains. Despite this success, however, it has been found that DNNs are surprisingly vulnerable to malicious attacks; adding a small, perceptually indistinguishable perturbations to the data can easily degrade classification performance. Adversarial training is an effective defense strategy to train a robust classifier. In this work, we propose to utilize the generator to learn how to create adversarial examples. Unlike the existing approaches that create a one-shot perturbation by a deterministic generator, we propose a recursive and stochastic generator that produces much stronger and diverse perturbations that comprehensively reveal the vulnerability of the target classifier. Our experiment results on MNIST and CIFAR-10 datasets show that the classifier adversarially trained with our method yields more robust performance over various white-box and black-box attacks.

Deep neural networks are vulnerable to adversarial attacks which can fool them by adding minuscule perturbations to the input images. The robustness of existing defenses suffers greatly under white-box attack settings, where an adversary has full knowledge about the network and can iterate several times to find strong perturbations. We observe that the main reason for the existence of such perturbations is the close proximity of different class samples in the learned feature space. This allows model decisions to be totally changed by adding an imperceptible perturbation in the inputs. To counter this, we propose to class-wise disentangle the intermediate feature representations of deep networks. Specifically, we force the features for each class to lie inside a convex polytope that is maximally separated from the polytopes of other classes. In this manner, the network is forced to learn distinct and distant decision regions for each class. We observe that this simple constraint on the features greatly enhances the robustness of learned models, even against the strongest white-box attacks, without degrading the classification performance on clean images. We report extensive evaluations in both black-box and white-box attack scenarios and show significant gains in comparison to state-of-the art defenses.

Deep neural networks (DNNs) have been widely applied in various applications, including autonomous driving and surveillance systems. However, DNNs are found to be vulnerable to adversarial examples, which are carefully crafted inputs aiming to mislead a learner to make incorrect predictions. While several defense and detection approaches are proposed for static image classification, many security-critical tasks use videos as their input and require efficient processing. In this paper, we propose an efficient and effective method advIT to detect adversarial frames within videos against different types of attacks based on temporal consistency property of videos. In particular, we apply optical flow estimation to the target and previous frames to generate pseudo frames and evaluate the consistency of the learner output between these pseudo frames and target. High inconsistency indicates that the target frame is adversarial. We conduct extensive experiments on various learning tasks including video semantic segmentation, human pose estimation, object detection, and action recognition, and demonstrate that we can achieve above 95% adversarial frame detection rate. To consider adaptive attackers, we show that even if an adversary has access to the detector and performs a strong adaptive attack based on the state of the art expectation of transformation method, the detection rate stays almost the same. We also tested the transferability among different optical flow estimators and show that it is hard for attackers to attack one and transfer the perturbation to others. In addition, as efficiency is important in video analysis, we show that advIT can achieve real-time detection in about 0.03--0.4 seconds.

We are concerned with the vulnerability of computer vision models to distributional shifts. We formulate a combinatorial optimization problem that allows evaluating the regions in the image space where a given model is more vulnerable, in terms of image transformations applied to the input, and face it with standard search algorithms. We further embed this idea in a training procedure, where we define new data augmentation rules according to the image transformations that the current model is most vulnerable to, over iterations. An empirical evaluation on classification and semantic segmentation problems suggests that the devised algorithm allows to train models that are more robust against content-preserving image manipulations and, in general, against distributional shifts.

Heatmap regression with a deep network has become one of the mainstream approaches to localize facial landmarks. However, the loss function for heatmap regression is rarely studied. In this paper, we analyze the ideal loss function properties for heatmap regression in face alignment problems. Then we propose a novel loss function, named Adaptive Wing loss, that is able to adapt its shape to different types of ground truth heatmap pixels. This adaptability penalizes loss more on foreground pixels while less on background pixels. To address the imbalance between foreground and background pixels, we also propose Weighted Loss Map, which assigns high weights on foreground and difficult background pixels to help training process focus more on pixels that are crucial to landmark localization. To further improve face alignment accuracy, we introduce boundary prediction and CoordConv with boundary coordinates. Extensive experiments on different benchmarks, including COFW, 300W and WFLW, show our approach outperforms the state-of-the-art by a significant margin on various evaluation metrics. Besides, the Adaptive Wing loss also helps other heatmap regression tasks.

Weakly supervised referring expression grounding aims at localizing the referential object in an image according to the linguistic query, where the mapping between the referential object and query is unknown in the training stage. To address this problem, we propose a novel end-to-end adaptive reconstruction network (ARN). It builds the correspondence between image region proposal and query in an adaptive manner: adaptive grounding and collaborative reconstruction. Specifically, we first extract the subject, location and context features to represent the proposals and the query respectively. Then, we design the adaptive grounding module to compute the matching score between each proposal and query by a hierarchical attention model. Finally, based on attention score and proposal features, we reconstruct the input query with a collaborative loss of language reconstruction loss, adaptive reconstruction loss, and attribute classification loss. This adaptive mechanism helps our model to alleviate the variance of different referring expressions. Experiments on four large-scale datasets show ARN outperforms existing state-of-the-art methods by a large margin. Qualitative results demonstrate that the proposed ARN can better handle the situation where multiple objects of a particular category situated together.

Crowd counting is an important topic in computer vision due to its practical usage in surveillance systems. The typical design of crowd counting algorithms is divided into two steps. First, the ground-truth density maps of crowd images are generated from the ground-truth dot maps (density map generation), e.g., by convolving with a Gaussian kernel. Second, deep learning models are designed to predict a density map from an input image (density map estimation). Most research efforts have concentrated on the density map estimation problem, while the problem of density map generation has not been adequately explored. In particular, the density map could be considered as an intermediate representation used to train a crowd counting network. In the sense of end-to-end training, the hand-crafted methods used for generating the density maps may not be optimal for the particular network or dataset used. To address this issue, we first show the impact of different density maps and that better ground-truth density maps can be obtained by refining the existing ones using a learned refinement network, which is jointly trained with the counter. Then, we propose an adaptive density map generator, which takes the annotation dot map as input, and learns a density map representation for a counter. The counter and generator are trained jointly within an end-to-end framework. The experiment results on popular counting datasets confirm the effectiveness of the proposed learnable density map representations.

Recent works attempt to improve scene parsing performance by exploring different levels of contexts, and typically train a well-designed convolutional network to exploit useful contexts across all pixels equally. However, in this paper, we find that the context demands are varying from different pixels or regions in each image. Based on this observation, we propose an Adaptive Context Network (ACNet) to capture the pixel-aware contexts by a competitive fusion of global context and local context according to different per-pixel demands. Specifically, when given a pixel, the global context demand is measured by the similarity between the global feature and its local feature, whose reverse value can also be used to measure the local context demand. We model the two demanding measurements by the proposed global context module and local context module, respectively, to generate their adaptive contextual features. Furthermore, we import multiple such modules to build several adaptive context blocks in different levels of network to obtain a coarse-to-fine result. Finally, comprehensive experimental evaluations demonstrate the effectiveness of the proposed ACNet, and new state-of-the-arts performances are achieved on all four public datasets, i.e. Cityscapes, ADE20K, PASCAL Context, and COCO Stuff.

There is a growing interest in strategies that can help us understand or interpret neural networks -- that is, not merely provide a prediction, but also offer additional context explaining why and how. While many current methods offer tools to perform this analysis for a given (trained) network post-hoc, recent results (especially on capsule networks) suggest that when classes map to a few high level "concepts" in the preceding layers of the network, the behavior of the network is easier to interpret or explain. Such training may be accomplished via dynamic/EM routing where the network "routes" for individual classes (or subsets of images) are dynamic and involve few nodes even if the full network may not be sparse. In this paper, we show how a simple modification of the SGD scheme can help provide dynamic/EM routing type behavior in convolutional neural networks. Through extensive experiments, we evaluate the effect of this idea for interpretability where we obtain promising results, while also showing that no compromise in attainable accuracy is involved. Further, we show that the minor modification is seemingly ad-hoc, the new algorithm can be analyzed by an approximate method which provably matches known rates for SGD.

Despite the outstanding performance of convolutional neural networks (CNNs) for many vision tasks, the required computational cost during inference is problematic when resources are limited. In this context, we propose Convolutional Neural Mixture Models (CNMMs), a probabilistic model embedding a large number of CNNs that can be jointly trained and evaluated in an efficient manner. Within the proposed framework, we present different mechanisms to prune subsets of CNNs from the mixture, allowing to easily adapt the computational cost required for inference. Image classification and semantic segmentation experiments show that our method achieve excellent accuracy-compute trade-offs. Moreover, unlike most of previous approaches, a single CNMM provides a large range of operating points along this trade-off, without any re-training.