CVPR 2018 论文列表

We present DeepMVS, a deep convolutional neural network (ConvNet) for multi-view stereo reconstruction. Taking an arbitrary number of posed images as input, we first produce a set of plane-sweep volumes and use the proposed DeepMVS network to predict high-quality disparity maps. The key contributions that enable these results are (1) supervised pretraining on a photorealistic synthetic dataset, (2) an effective method for aggregating information across a set of unordered images, and (3) integrating multi-layer feature activations from the pre-trained VGG-19 network. We validate the efficacy of DeepMVS using the ETH3D Benchmark. Our results show that DeepMVS compares favorably against state-of-the-art conventional MVS algorithms and other ConvNet based methods, particularly for near-textureless regions and thin structures.

Video super-resolution (VSR) has become even more important recently to provide high resolution (HR) contents for ultra high definition displays. While many deep learning based VSR methods have been proposed, most of them rely heavily on the accuracy of motion estimation and compensation. We introduce a fundamentally different framework for VSR in this paper. We propose a novel end-to-end deep neural network that generates dynamic upsampling filters and a residual image, which are computed depending on the local spatio-temporal neighborhood of each pixel to avoid explicit motion compensation. With our approach, an HR image is reconstructed directly from the input image using the dynamic upsampling filters, and the fine details are added through the computed residual. Our network with the help of a new data augmentation technique can generate much sharper HR videos with temporal consistency, compared with the previous methods. We also provide analysis of our network through extensive experiments to show how the network deals with motions implicitly.

The success of current deep saliency detection methods heavily depends on the availability of large-scale supervision in the form of per-pixel labeling. Such supervision, while labor-intensive and not always possible, tends to hinder the generalization ability of the learned models. By contrast, traditional handcrafted features based unsupervised saliency detection methods, even though have been surpassed by the deep supervised methods, are generally dataset-independent and could be applied in the wild. This raises a natural question that ``Is it possible to learn saliency maps without using labeled data while improving the generalization ability?''. To this end, we present a novel perspective to unsupervised saliency detection through learning from multiple noisy labeling generated by ``weak'' and ``noisy'' unsupervised handcrafted saliency methods. Our end-to-end deep learning framework for unsupervised saliency detection consists of a latent saliency prediction module and a noise modeling module that work collaboratively and are optimized jointly. Explicit noise modeling enables us to deal with noisy saliency maps in a probabilistic way. Extensive experimental results on various benchmarking datasets show that our model not only outperforms all the unsupervised saliency methods with a large margin but also achieves comparable performance with the recent state-of-the-art supervised deep saliency methods.

We present a texture network called Deep Encoding Pooling Network (DEP) for the task of ground terrain recognition. Recognition of ground terrain is an important task in establishing robot or vehicular control parameters, as well as for localization within an outdoor environment. The architecture of DEP integrates orderless texture details and local spatial information and the performance of DEP surpasses state-of-the-art methods for this task. The GTOS database (comprised of over 30,000 images of 40 classes of ground terrain in outdoor scenes) enables supervised recognition. For evaluation under realistic conditions, we use test images that are not from the existing GTOS dataset, but are instead from hand-held mobile phone videos of similar terrain. This new evaluation dataset, GTOS-mobile, consists of 81 videos of 31 classes of ground terrain such as grass, gravel, asphalt and sand. The resultant network shows excellent performance not only for GTOS-mobile, but also for more general databases (MINC and DTD). Leveraging the discriminant features learned from this network, we build a new texture manifold called DEP-manifold. We learn a parametric distribution in feature space in a fully supervised manner, which gives the distance relationship among classes and provides a means to implicitly represent ambiguous class boundaries. The source code and database are publicly available.

In this work we introduce a time- and memory-efficient method for structured prediction that couples neuron decisions across both space at time. We show that we are able to perform exact and efficient inference on a densely connected spatio-temporal graph by capitalizing on recent advances on deep Gaussian Conditional Random Fields (GCRFs). Our method, called VideoGCRF is (a) efficient, (b) has a unique global minimum, and (c) can be trained end-to-end alongside contemporary deep networks for video understanding. We experiment with multiple connectivity patterns in the temporal domain, and present empirical improvements over strong baselines on the tasks of both semantic and instance segmentation of videos. Our implementation is based on the Caffe2 framework and will be available at https://github.com/siddharthachandra/gcrf-v3.0.

Partial person re-identification (re-id) is a challenging problem, where only a partial observation of a person image is available for matching. However, few studies have offered a solution of how to identify an arbitrary patch of a person image. In this paper, we propose a fast and accurate matching method to address this problem. The proposed method leverages Fully Convolutional Network (FCN) to generate correspondingly-size spatial feature maps such that pixel-level features are consistent. To match a pair of person images of different sizes, a novel method called Deep Spatial feature Reconstruction (DSR) is further developed to avoid explicit alignment. Specifically, we exploit the reconstructing error from dictionary learning to calculate the similarity between different spatial feature maps. In that way, we expect that the proposed FCN can decrease the similarity of coupled images from different persons and vice versa. Experimental results on two partial person datasets demonstrate the efficiency and effectiveness of the proposed method in comparison with several state-of-the-art partial person re-id approaches.

Deep feed-forward convolutional neural networks (CNNs) have become ubiquitous in virtually all machine learning and computer vision challenges; however, advancements in CNNs have arguably reached an engineering saturation point where incremental novelty results in minor performance gains. Although there is evidence that object classification has reached human levels on narrowly defined tasks, for general applications, the biological visual system is far superior to that of any computer. Research reveals there are numerous missing components in feed-forward deep neural networks that are critical in mammalian vision. The brain does not work solely in a feed-forward fashion, but rather all of the neurons are in competition with each other; neurons are integrating information in a bottom up and top down fashion and incorporating expectation and feedback in the modeling process. Furthermore, our visual cortex is working in tandem with our parietal lobe, integrating sensory information from various modalities. In our work, we sought to improve upon the standard feed-forward deep learning model by augmenting them with biologically inspired concepts of sparsity, top down feedback, and lateral inhibition. We define our model as a sparse coding problem using hierarchical layers. We solve the sparse coding problem with an additional top down feedback error driving the dynamics of the neural network. While building and observing the behavior of our model, we were fascinated that multimodal, invariant neurons naturally emerged that mimicked, "Halle Berry neurons" found in the human brain. These neurons trained in our sparse model learned to respond to high level concepts from multiple modalities, which is not the case with a standard feed-forward autoencoder. Furthermore, our sparse representation of multimodal signals demonstrates qualitative and quantitative superiority to the standard feed-forward joint embedding in common vision and machine learning tasks.

In this paper, we present an effective and efficient face deblurring algorithm by exploiting semantic cues via deep convolutional neural networks (CNNs). As face images are highly structured and share several key semantic components (e.g., eyes and mouths), the semantic information of a face provides a strong prior for restoration. As such, we propose to incorporate global semantic priors as input and impose local structure losses to regularize the output within a multi-scale deep CNN. We train the network with perceptual and adversarial losses to generate photo-realistic results and develop an incremental training strategy to handle random blur kernels in the wild. Quantitative and qualitative evaluations demonstrate that the proposed face deblurring algorithm restores sharp images with more facial details and performs favorably against state-of-the-art methods in terms of restoration quality, face recognition and execution speed.

We propose drl-RPN, a deep reinforcement learning-based visual recognition model consisting of a sequential region proposal network (RPN) and an object detector. In contrast to typical RPNs, where candidate object regions (RoIs) are selected greedily via class-agnostic NMS, drl-RPN optimizes an objective closer to the final detection task. This is achieved by replacing the greedy RoI selection process with a sequential attention mechanism which is trained via deep reinforcement learning (RL). Our model is capable of accumulating class-specific evidence over time, potentially affecting subsequent proposals and classification scores, and we show that such context integration significantly boosts detection accuracy. Moreover, drl-RPN automatically decides when to stop the search process and has the benefit of being able to jointly learn the parameters of the policy and the detector, both represented as deep networks. Our model can further learn to search over a wide range of exploration-accuracy trade-offs making it possible to specify or adapt the exploration extent at test time. The resulting search trajectories are image- and category-dependent, yet rely only on a single policy over all object categories. Results on the MS COCO and PASCAL VOC challenges show that our approach outperforms established, typical state-of-the-art object detection pipelines.

Age estimation from facial images is typically cast as a nonlinear regression problem. The main challenge of this problem is the facial feature space w.r.t. ages is inhomogeneous, due to the large variation in facial appearance across different persons of the same age and the non-stationary property of aging patterns. In this paper, we propose Deep Regression Forests (DRFs), an end-to-end model, for age estimation. DRFs connect the split nodes to a fully connected layer of a convolutional neural network (CNN) and deal with inhomogeneous data by jointly learning input-dependant data partitions at the split nodes and data abstractions at the leaf nodes. This joint learning follows an alternating strategy: First, by fixing the leaf nodes, the split nodes as well as the CNN parameters are optimized by Back-propagation; Then, by fixing the split nodes, the leaf nodes are optimized by iterating a step-size free update rule derived from Variational Bounding. We verify the proposed DRFs on three standard age estimation benchmarks and achieve state-of-the-art results on all of them.

In this paper, we propose a deep progressive reinforcement learning (DPRL) method for action recognition in skeleton-based videos, which aims to distil the most informative frames and discard ambiguous frames in sequences for recognizing actions. Since the choices of selecting representative frames are multitudinous for each video, we model the frame selection as a progressive process through deep reinforcement learning, during which we progressively adjust the chosen frames by taking two important factors into account: (1) the quality of the selected frames and (2) the relationship between the selected frames to the whole video. Moreover, considering the topology of human body inherently lies in a graph-based structure, where the vertices and edges represent the hinged joints and rigid bones respectively, we employ the graph-based convolutional neural network to capture the dependency between the joints for action recognition. Our approach achieves very competitive performance on three widely used benchmarks.

This paper proposes an unpaired learning method for image enhancement. Given a set of photographs with the desired characteristics, the proposed method learns a photo enhancer which transforms an input image into an enhanced image with those characteristics. The method is based on the framework of two-way generative adversarial networks (GANs) with several improvements. First, we augment the U-Net with global features and show that it is more effective. The global U-Net acts as the generator in our GAN model. Second, we improve Wasserstein GAN (WGAN) with an adaptive weighting scheme. With this scheme, training converges faster and better, and is less sensitive to parameters than WGAN-GP. Finally, we propose to use individual batch normalization layers for generators in two-way GANs. It helps generators better adapt to their own input distributions. All together, they significantly improve the stability of GAN training for our application. Both quantitative and visual results show that the proposed method is effective for enhancing images.

Standard convolutional neural networks assume a grid structured input is available and exploit discrete convolutions as their fundamental building blocks. This limits their applicability to many real-world applications. In this paper we propose Parametric Continuous Convolution, a new learnable operator that operates over non-grid structured data. The key idea is to exploit parameterized kernel functions that span the full continuous vector space. This generalization allows us to learn over arbitrary data structures as long as their support relationship is computable. Our experiments show significant improvement over the state-of-the-art in point cloud segmentation of indoor and outdoor scenes, and lidar motion estimation of driving scenes.

Monocular depth estimation, which plays a crucial role in understanding 3D scene geometry, is an ill-posed prob- lem. Recent methods have gained significant improvement by exploring image-level information and hierarchical features from deep convolutional neural networks (DCNNs). These methods model depth estimation as a regression problem and train the regression networks by minimizing mean squared error, which suffers from slow convergence and unsatisfactory local solutions. Besides, existing depth estimation networks employ repeated spatial pooling operations, resulting in undesirable low-resolution feature maps. To obtain high-resolution depth maps, skip-connections or multi- layer deconvolution networks are required, which complicates network training and consumes much more computations. To eliminate or at least largely reduce these problems, we introduce a spacing-increasing discretization (SID) strategy to discretize depth and recast depth network learning as an ordinal regression problem. By training the network using an ordinary regression loss, our method achieves much higher accuracy and faster convergence in synch. Furthermore, we adopt a multi-scale network structure which avoids unnecessary spatial pooling and captures multi-scale information in parallel. The proposed deep ordinal regression network (DORN) achieves state-of-the-art results on three challenging benchmarks, i.e., KITTI [16], Make3D [49], and NYU Depth v2 [41], and outperforms existing methods by a large margin.

Model distillation is an effective and widely used technique to transfer knowledge from a teacher to a student network. The typical application is to transfer from a powerful large network or ensemble to a small network, in order to meet the low-memory or fast execution requirements. In this paper, we present a deep mutual learning (DML) strategy. Different from the one-way transfer between a static pre-defined teacher and a student in model distillation, with DML, an ensemble of students learn collaboratively and teach each other throughout the training process. Our experiments show that a variety of network architectures benefit from mutual learning and achieve compelling results on both category and instance recognition tasks. Surprisingly, it is revealed that no prior powerful teacher network is necessary -- mutual learning of a collection of simple student networks works, and moreover outperforms distillation from a more powerful yet static teacher.

Cross-spectral imaging provides strong benefits for recognition and detection tasks. Often, multiple cameras are used for cross-spectral imaging, thus requiring image alignment, or disparity estimation in a stereo setting. Increasingly, multi-camera cross-spectral systems are embedded in active RGBD devices (e.g. RGB-NIR cameras in Kinect and iPhone X). Hence, stereo matching also provides an opportunity to obtain depth without an active projector source. However, matching images from different spectral bands is challenging because of large appearance variations. We develop a novel deep learning framework to simultaneously transform images across spectral bands and estimate disparity. A material-aware loss function is incorporated within the disparity prediction network to handle regions with unreliable matching such as light sources, glass windshields and glossy surfaces. No depth supervision is required by our method. To evaluate our method, we used a vehicle-mounted RGB-NIR stereo system to collect 13.7 hours of video data across a range of areas in and around a city. Experiments show that our method achieves strong performance and reaches real-time speed.

Existing learning based solutions to 3D surface prediction cannot be trained end-to-end as they operate on intermediate representations (e.g., TSDF) from which 3D surface meshes must be extracted in a post-processing step (e.g., via the marching cubes algorithm). In this paper, we investigate the problem of end-to-end 3D surface prediction. We first demonstrate that the marching cubes algorithm is not differentiable and propose an alternative differentiable formulation which we insert as a final layer into a 3D convolutional neural network. We further propose a set of loss functions which allow for training our model with sparse point supervision. Our experiments demonstrate that the model allows for predicting sub-voxel accurate 3D shapes of arbitrary topology. Additionally, it learns to complete shapes and to separate an object's inside from its outside even in the presence of sparse and incomplete ground truth. We investigate the benefits of our approach on the task of inferring shapes from 3D point clouds. Our model is flexible and can be combined with a variety of shape encoder and shape inference techniques.

Radiologists in their daily work routinely find and annotate significant abnormalities on a large number of radiology images. Such abnormalities, or lesions, have collected over years and stored in hospitals' picture archiving and communication systems. However, they are basically unsorted and lack semantic annotations like type and location. In this paper, we aim to organize and explore them by learning a deep feature representation for each lesion. A large-scale and comprehensive dataset, DeepLesion, is introduced for this task. DeepLesion contains bounding boxes and size measurements of over 32K lesions. To model their similarity relationship, we leverage multiple supervision information including types, self-supervised location coordinates, and sizes. They require little manual annotation effort but describe useful attributes of the lesions. Then, a triplet network is utilized to learn lesion embeddings with a sequential sampling strategy to depict their hierarchical similarity structure. Experiments show promising qualitative and quantitative results on lesion retrieval, clustering, and classification. The learned embeddings can be further employed to build a lesion graph for various clinically useful applications. An algorithm for intra-patient lesion matching is proposed and validated with experiments.

The problem of graph matching under node and pair-wise constraints is fundamental in areas as diverse as combinatorial optimization, machine learning or computer vision, where representing both the relations between nodes and their neighborhood structure is essential. We present an end-to-end model that makes it possible to learn all parameters of the graph matching process, including the unary and pairwise node neighborhoods, represented as deep feature extraction hierarchies. The challenge is in the formulation of the different matrix computation layers of the model in a way that enables the consistent, efficient propagation of gradients in the complete pipeline from the loss function, through the combinatorial optimization layer solving the matching problem, and the feature extraction hierarchy. Our computer vision experiments and ablation studies on challenging datasets like PASCAL VOC keypoints, Sintel and CUB show that matching models refined end-to-end are superior to counterparts based on feature hierarchies trained for other problems.