CVPR 2019 Papers

Automated Machine Learning (AutoML) is an active area on the design of deep neural networks for specific tasks and datasets. Given the complexity of discovering new network designs, methods for speeding up the search procedure are becoming important. This paper presents a so-called transferable AutoML approach that Automated Machine Learning (AutoML) is an active area on the design of deep neural networks for specific tasks and datasets. Given the complexity of discovering new network designs, methods for speeding up the search procedure are becoming important. This paper presents a so-called transferable AutoML approach that leverages previously trained models to speed up the search process for new tasks and datasets. Our approach involves a novel meta-feature extraction technique based on the performance of benchmark models, and a dynamic dataset clustering algorithm based on Markov process and statistical hypothesis test. As such multiple models can share a common structure while with different learned parameters. The transferable AutoML can either be applied to search from scratch, search from predesigned models, or transfer from basic cells according to the difficulties of the given datasets. The experimental results on image classification show notable speedup in overall search time for multiple datasets with negligible loss in accuracy.

We study how to leverage off-the-shelf visual and linguistic data to cope with out-of-vocabulary answers in visual question answering task. Existing large-scale visual datasets with annotations such as image class labels, bounding boxes and region descriptions are good sources for learning rich and diverse visual concepts. However, it is not straightforward how the visual concepts can be captured and transferred to visual question answering models due to missing link between question dependent answering models and visual data without question. We tackle this problem in two steps: 1) learning a task conditional visual classifier, which is capable of solving diverse question-specific visual recognition tasks, based on unsupervised task discovery and 2) transferring the task conditional visual classifier to visual question answering models. Specifically, we employ linguistic knowledge sources such as structured lexical database (e.g. WordNet) and visual descriptions for unsupervised task discovery, and transfer a learned task conditional visual classifier as an answering unit in a visual question answering model. We empirically show that the proposed algorithm generalizes to out-of-vocabulary answers successfully using the knowledge transferred from the visual dataset.

Unsupervised image-to-image translation aims at learning a mapping between two visual domains. However, learning a translation across large geometry variations al- ways ends up with failure. In this work, we present a novel disentangle-and-translate framework to tackle the complex objects image-to-image translation task. Instead of learning the mapping on the image space directly, we disentangle image space into a Cartesian product of the appearance and the geometry latent spaces. Specifically, we first in- troduce a geometry prior loss and a conditional VAE loss to encourage the network to learn independent but com- plementary representations. The translation is then built on appearance and geometry space separately. Extensive experiments demonstrate the superior performance of our method to other state-of-the-art approaches, especially in the challenging near-rigid and non-rigid objects translation tasks. In addition, by taking different exemplars as the ap- pearance references, our method also supports multimodal translation. Project page: https://wywu.github. io/projects/TGaGa/TGaGa.html

Compressive sensing is a method to recover the original image from undersampled measurements. In order to overcome the ill-posedness of this inverse problem, image priors are used such as sparsity, minimal total-variation, or self-similarity of images. Recently, deep learning based compressive image recovery methods have been proposed and have yielded state-of-the-art performances. They used data-driven approaches instead of hand-crafted image priors to regularize ill-posed inverse problems with undersampled data. Ironically, training deep neural networks (DNNs) for them requires "clean" ground truth images, but obtaining the best quality images from undersampled data requires well-trained DNNs. To resolve this dilemma, we propose novel methods based on two well-grounded theories: denoiser-approximate message passing (D-AMP) and Stein's unbiased risk estimator (SURE). Our proposed methods were able to train deep learning based image denoisers from undersampled measurements without ground truth images and without additional image priors, and to recover images with state-of-the-art qualities from undersampled data. We evaluated our methods for various compressive sensing recovery problems with Gaussian random, coded diffraction pattern, and compressive sensing MRI measurement matrices. Our proposed methods yielded state-of-the-art performances for all cases without ground truth images. Our methods also yielded comparable performances to the methods with ground truth data.

Online Multi-Object Tracking (MOT) from videos is a challenging computer vision task which has been extensively studied for decades. Most of the existing MOT algorithms are based on the Tracking-by-Detection (TBD) paradigm combined with popular machine learning approaches which largely reduce the human effort to tune algorithm parameters. However, the commonly used supervised learning approaches require the labeled data (e.g., bounding boxes), which is expensive for videos. Also, the TBD framework is usually suboptimal since it is not end-to-end, i.e., it considers the task as detection and tracking, but not jointly. To achieve both label-free and end-to-end learning of MOT, we propose a Tracking-by-Animation framework, where a differentiable neural model first tracks objects from input frames and then animates these objects into reconstructed frames. Learning is then driven by the reconstruction error through backpropagation. We further propose a Reprioritized Attentive Tracking to improve the robustness of data association. Experiments conducted on both synthetic and real video datasets show the potential of the proposed model. Our project page is publicly available at: https://github.com/zhen-he/tracking-by-animation

Interest in image-to-image translation has grown substantially in recent years with the success of unsupervised models based on the cycle-consistency assumption. The achievements of these models have been limited to a particular subset of domains where this assumption yields good results, namely homogeneous domains that are characterized by style or texture differences. We tackle the challenging problem of image-to-image translation where the domains are defined by high-level shapes and contexts, as well as including significant clutter and heterogeneity. For this purpose, we introduce a novel GAN based on preserving intra-domain vector transformations in a latent space learned by a siamese network. The traditional GAN system introduced a discriminator network to guide the generator into generating images in the target domain. To this two-network system we add a third: a siamese network that guides the generator so that each original image shares semantics with its generated version. With this new three-network system, we no longer need to constrain the generators with the ubiquitous cycle-consistency restraint or any other autoencoding regularization. As a result, the generators can learn mappings between more complex domains that differ from each other by more than just style or texture. We demonstrate our model by mapping between high-resolution, arbitrarily chosen classes from the Imagenet dataset completely without pre-processing such as cropping, centering, or filtering unrepresentative images.

We present a new algorithm for predicting the near-term trajectories of road agents in dense traffic videos. Our approach is designed for heterogeneous traffic, where the road agents may correspond to buses, cars, scooters, bi-cycles, or pedestrians. We model the interactions between different road agents using a novel LSTM-CNN hybrid network for trajectory prediction. In particular, we take into account heterogeneous interactions that implicitly account for the varying shapes, dynamics, and behaviors of different road agents. In addition, we model horizon-based interactions which are used to implicitly model the driving behavior of each road agent. We evaluate the performance of our prediction algorithm, TraPHic, on the standard datasets and also introduce a new dense, heterogeneous traffic dataset corresponding to urban Asian videos and agent trajectories. We outperform state-of-the-art methods on dense traffic datasets by 30%.

Most existing visual search systems are deployed based upon fixed kinds of visual features, which prohibits the feature reusing across different systems or when upgrading systems with a new type of feature. Such a setting is obviously inflexible and time/memory consuming, which is indeed mendable if visual features can be "translated" across systems. In this paper, we make the first attempt towards visual feature translation to break through the barrier of using features across different visual search systems. To this end, we propose a Hybrid Auto-Encoder (HAE) to translate visual features, which learns a mapping by minimizing the translation and reconstruction errors. Based upon HAE, an Undirected Affinity Measurement (UAM) is further designed to quantify the affinity among different types of visual features. Extensive experiments have been conducted on several public datasets with sixteen different types of widely-used features in visual search systems. Quantitative results show the encouraging possibilities of feature translation. For the first time, the affinity among widely-used features like SIFT and DELF is reported.

Studies have shown that a dominant class of questions asked by visually impaired users on images of their surroundings involves reading text in the image. But today's VQA models can not read! Our paper takes a first step towards addressing this problem. First, we introduce a new "TextVQA" dataset to facilitate progress on this important problem. Existing datasets either have a small proportion of questions about text (e.g., the VQA dataset) or are too small (e.g., the VizWiz dataset). TextVQA contains 45,336 questions on 28,408 images that require reasoning about text to answer. Second, we introduce a novel model architecture that reads text in the image, reasons about it in the context of the image and the question, and predicts an answer which might be a deduction based on the text and the image or composed of the strings found in the image. Consequently, we call our approach Look, Read, Reason & Answer (LoRRA). We show that LoRRA outperforms existing state-of-the-art VQA models on our TextVQA dataset. We find that the gap between human performance and machine performance is significantly larger on TextVQA than on VQA 2.0, suggesting that TextVQA is well-suited to benchmark progress along directions complementary to VQA 2.0.

Despite increasing efforts on universal representations for visual recognition, few have addressed object detection. In this paper, we develop an effective and efficient universal object detection system that is capable of working on various image domains, from human faces and traffic signs to medical CT images. Unlike multi-domain models, this universal model does not require prior knowledge of the domain of interest. This is achieved by the introduction of a new family of adaptation layers, based on the principles of squeeze and excitation, and a new domain-attention mechanism. In the proposed universal detector, all parameters and computations are shared across domains, and a single network processes all domains all the time. Experiments, on a newly established universal object detection benchmark of 11 diverse datasets, show that the proposed detector outperforms a bank of individual detectors, a multi-domain detector, and a baseline universal detector, with a 1.3x parameter increase over a single-domain baseline detector. The code and benchmark are available at http://www.svcl.ucsd.edu/projects/universal-detection/.

We present a new research task and a dataset to understand human social interactions via computational methods, to ultimately endow machines with the ability to encode and decode a broad channel of social signals humans use. This research direction is essential to make a machine that genuinely communicates with humans, which we call Social Artificial Intelligence. We first formulate the "social signal prediction" problem as a way to model the dynamics of social signals exchanged among interacting individuals in a data-driven way. We then present a new 3D motion capture dataset to explore this problem, where the broad spectrum of social signals (3D body, face, and hand motions) are captured in a triadic social interaction scenario. Baseline approaches to predict speaking status, social formation, and body gestures of interacting individuals are presented in the defined social prediction framework.

Monocular depth estimation is a challenging task in scene understanding, with the goal to acquire the geometric properties of 3D space from 2D images. Due to the lack of RGB-depth image pairs, unsupervised learning methods aim at deriving depth information with alternative supervision such as stereo pairs. However, most existing works fail to model the geometric structure of objects, which generally results from considering pixel-level objective functions during training. In this paper, we propose SceneNet to overcome this limitation with the aid of semantic understanding from segmentation. Moreover, our proposed model is able to perform region-aware depth estimation by enforcing semantics consistency between stereo pairs. In our experiments, we qualitatively and quantitatively verify the effectiveness and robustness of our model, which produces favorable results against the state-of-the-art approaches do.

It is challenging to detect curve texts due to their irregular shapes and varying sizes. In this paper, we first investigate the deficiency of the existing curve detection methods and then propose a novel Conditional Spatial Expansion (CSE) mechanism to improve the performance of curve detection. Instead of regarding the curve text detection as a polygon regression or a segmentation problem, we formulate it as a sequence prediction on the spatial domain. CSE starts with a seed arbitrarily chosen within a text region and progressively merges neighborhood regions based on the extracted local features by a CNN and contextual information of merged regions. The CSE is highly parameterized and can be seamlessly integrated into existing object detection frameworks. Enhanced by the data-dependent CSE mechanism, our curve text detection system provides robust instance-level text region extraction with minimal post-processing. The analysis experiment shows that our CSE can handle texts with various shapes, sizes, and orientations, and can effectively suppress the false-positives coming from text-like textures or unexpected texts included in the same RoI. Compared with the existing curve text detection algorithms, our method is more robust and enjoys a simpler processing flow. It also creates a new state-of-art performance on curve text benchmarks with F-measurement of up to 78.4%.

The fundamental challenge of small inter-person variation requires Person Re-Identification (Re-ID) models to capture sufficient fine-grained information. This paper proposes to discover diverse discriminative visual cues without extra assistance, e.g., pose estimation, human parsing. Specifically, a Class Activation Maps (CAM) augmentation model is proposed to expand the activation scope of baseline Re-ID model to explore rich visual cues, where the backbone network is extended by a series of ordered branches which share the same input but output complementary CAM. A novel Overlapped Activation Penalty is proposed to force the new branch to pay more attention to the image regions less activated by the old ones, such that spatial diverse visual features can be discovered. The proposed model achieves state-of-the-art results on three person Re-ID benchmarks. Moreover, a visualization approach termed ranking activation map (RAM) is proposed to explicitly interpret the ranking results in the test stage, which gives qualitative validations of the proposed method.

Most existing super-resolution methods do not perform well in real scenarios due to lack of realistic training data and information loss of the model input. To solve the first problem, we propose a new pipeline to generate realistic training data by simulating the imaging process of digital cameras. And to remedy the information loss of the input, we develop a dual convolutional neural network to exploit the originally captured radiance information in raw images. In addition, we propose to learn a spatially-variant color transformation which helps more effective color corrections. Extensive experiments demonstrate that super-resolution with raw data helps recover fine details and clear structures, and more importantly, the proposed network and data generation pipeline achieve superior results for single image super-resolution in real scenarios.

Structured pruning of filters or neurons has received increased focus for compressing convolutional neural networks. Most existing methods rely on multi-stage optimizations in a layer-wise manner for iteratively pruning and retraining which may not be optimal and may be computation intensive. Besides, these methods are designed for pruning a specific structure, such as filter or block structures without jointly pruning heterogeneous structures. In this paper, we propose an effective structured pruning approach that jointly prunes filters as well as other structures in an end-to-end manner. To accomplish this, we first introduce a soft mask to scale the output of these structures by defining a new objective function with sparsity regularization to align the output of baseline and network with this mask. We then effectively solve the optimization problem by generative adversarial learning (GAL), which learns a sparse soft mask in a label-free and an end-to-end manner. By forcing more scale factors in the soft mask to zero, the fast iterative shrinkage-thresholding algorithm (FISTA) can be leveraged to fast and reliably remove the corresponding structures. Extensive experiments demonstrate the effectiveness of GAL on different datasets, including MNIST, CIFAR-10 and ImageNet ILSVRC 2012. For example, on ImageNet ILSVRC 2012, the pruned ResNet-50 achieves 10.88% Top-5 error and results in a factor of 3.7x speedup. This significantly outperforms state-of-the-art methods.

Anticipating the future motions of 3D articulate objects is challenging due to its non-linear and highly stochastic nature. Current approaches typically represent the skeleton of an articulate object as a set of 3D joints, which unfortunately ignores the relationship between joints, and fails to encode fine-grained anatomical constraints. Moreover, conventional recurrent neural networks, such as LSTM and GRU, are employed to model motion contexts, which inherently have difficulties in capturing long-term dependencies. To address these problems, we propose to explicitly encode anatomical constraints by modeling their skeletons with a Lie algebra representation. Importantly, a hierarchical recurrent network structure is developed to simultaneously encodes local contexts of individual frames and global contexts of the sequence. We proceed to explore the applications of our approach to several distinct quantities including human, fish, and mouse. Extensive experiments show that our approach achieves more natural and accurate predictions over state-of-the-art methods.

Unpaired Image-to-image Translation is a new rising and challenging vision problem that aims to learn a mapping between unaligned image pairs in diverse domains. Recent advances in this field like MUNIT and DRIT mainly focus on disentangling content and style/attribute from a given image first, then directly adopting the global style to guide the model to synthesize new domain images. However, this kind of approaches severely incurs contradiction if the target domain images are content-rich with multiple discrepant objects. In this paper, we present a simple yet effective instance-aware image-to-image translation approach (INIT), which employs the fine-grained local (instance) and global styles to the target image spatially. The proposed INIT exhibits three import advantages: (1) the instance-level objective loss can help learn a more accurate reconstruction and incorporate diverse attributes of objects; (2) the styles used for target domain of local/global areas are from corresponding spatial regions in source domain, which intuitively is a more reasonable mapping; (3) the joint training process can benefit both fine and coarse granularity and incorporates instance information to improve the quality of global translation. We also collect a large-scale benchmark for the new instance-level translation task. We observe that our synthetic images can even benefit real-world vision tasks like generic object detection.

Embedding 3D morphable basis functions into deep neural networks opens great potential for models with better representation power. However, to faithfully learn those models from an image collection, it requires strong regularization to overcome ambiguities involved in the learning process. This critically prevents us from learning high fidelity face models which are needed to represent face images in high level of details. To address this problem, this paper presents a novel approach to learn additional proxies as means to side-step strong regularizations, as well as, leverages to promote detailed shape/albedo. To ease the learning, we also propose to use a dual-pathway network, a carefully-designed architecture that brings a balance between global and local-based models. By improving the nonlinear 3D morphable model in both learning objective and network architecture, we present a model which is superior in capturing higher level of details than the linear or its precedent nonlinear counterparts. As a result, our model achieves state-of-the-art performance on 3D face reconstruction by solely optimizing latent representations.