CVPR 2019 Papers

Pixel-accurate tracking of objects is a key element in many computer vision applications, often solved by iterated individual object tracking or instance segmentation followed by object matching. Here we introduce cross-classification clustering (3C), a technique that simultaneously tracks complex, interrelated objects in an image stack. The key idea in cross-classification is to efficiently turn a clustering problem into a classification problem by running a logarithmic number of independent classifications per image, letting the cross-labeling of these classifications uniquely classify each pixel to the object labels. We apply the 3C mechanism to achieve state-of-the-art accuracy in connectomics -- the nanoscale mapping of neural tissue from electron microscopy volumes. Our reconstruction system increases scalability by an order of magnitude over existing single-object tracking methods (such as flood-filling networks). This scalability is important for the deployment of connectomics pipelines, since currently the best performing techniques require computing infrastructures that are beyond the reach of most laboratories. Our algorithm may offer benefits in other domains that require pixel-accurate tracking of multiple objects, such as segmentation of videos and medical imagery.

We present a convolutional network architecture for direct feature learning on mesh surfaces through their atlases of texture maps. The texture map encodes the parameterization from 3D to 2D domain, rendering not only RGB values but also rasterized geometric features if necessary. Since the parameterization of texture map is not pre-determined, and depends on the surface topologies, we therefore introduce a novel cross-atlas convolution to recover the original mesh geodesic neighborhood, so as to achieve the invariance property to arbitrary parameterization. The proposed module is integrated into classification and segmentation architectures, which takes the input texture map of a mesh, and infers the output predictions. Our method not only shows competitive performances on classification and segmentation public benchmarks, but also paves the way for the broad mesh surfaces learning.

Regular model compression methods focus on RGB input. While cross domain tasks demand more DNN models, each domain often needs its own model. Consequently, for such tasks, the storage cost, memory footprint and computation cost increase dramatically compared to single RGB input. Moreover, the distinct appearance and special structure in cross domain tasks make it difficult to directly apply regular compression methods on it. In this paper, thus, we propose a new robust cross domain model compression method. Specifically, the proposed method compress cross domain models by structurally weight sharing, which is achieved by regularizing the models with graph embedding at training time. Due to the channel wise weights sharing, the proposed method can reduce computation cost without specially designed algorithm. In the experiments, the proposed method achieves state of the art results on two diverse tasks: action recognition and RGB-D scene recognition.

We present the Creative Flow+ Dataset, the first diverse multi-style artistic video dataset richly labeled with per-pixel optical flow, occlusions, correspondences, segmentation labels, normals, and depth. Our dataset includes 3000 animated sequences rendered using styles randomly selected from 40 textured line styles and 38 shading styles, spanning the range between flat cartoon fill and wildly sketchy shading. Our dataset includes 124K+ train set frames and 10K test set frames rendered at 1500x1500 resolution, far surpassing the largest available optical flow datasets in size. While modern techniques for tasks such as optical flow estimation achieve impressive performance on realistic images and video, today there is no way to gauge their performance on non-photorealistic images. Creative Flow+ poses a new challenge to generalize real-world Computer Vision to messy stylized content. We show that learning-based optical flow methods fail to generalize to this data and struggle to compete with classical approaches, and invite new research in this area. Our dataset and a new optical flow benchmark will be publicly available at: www.cs.toronto.edu/creativeflow/. We further release the complete dataset creation pipeline, allowing the community to generate and stylize their own data on demand.

Unsupervised domain adaptation algorithms aim to transfer the knowledge learned from one domain to another (e.g., synthetic to real images). The adapted representations often do not capture pixel-level domain shifts that are crucial for dense prediction tasks (e.g., semantic segmentation). In this paper, we present a novel pixel-wise adversarial domain adaptation algorithm. By leveraging image-to-image translation methods for data augmentation, our key insight is that while the translated images between domains may differ in styles, their predictions for the task should be consistent. We exploit this property and introduce a cross-domain consistency loss that enforces our adapted model to produce consistent predictions. Through extensive experimental results, we show that our method compares favorably against the state-of-the-art on a wide variety of unsupervised domain adaptation tasks.

We present a coordinate-free description of Carlsson-Weinshall duality between scene points and camera pinholes and use it to derive a new characterization of primal/dual multi-view geometry. In the case of three views, a particular set of reduced trilinearities provide a novel parameterization of camera geometry that, unlike existing ones, is subject only to very simple internal constraints. These trilinearities lead to new "quasi-linear" algorithms for primal and dual structure from motion. We include some preliminary experiments with real and synthetic data.

We present a new deep learning approach to pose-guided resynthesis of human photographs. At the heart of the new approach is the estimation of the complete body surface texture based on a single photograph. Since the input photograph always observes only a part of the surface, we suggest a new inpainting method that completes the texture of the human body. Rather than working directly with colors of texture elements, the inpainting network estimates an appropriate source location in the input image for each element of the body surface. This correspondence field between the input image and the texture is then further warped into the target image coordinate frame based on the desired pose, effectively establishing the correspondence between the source and the target view even when the pose change is drastic. The final convolutional network then uses the established correspondence and all other available information to synthesize the output image. A fully-convolutional architecture with deformable skip connections guided by the estimated correspondence field is used. We show state-of-the-art result for pose-guided image synthesis. Additionally, we demonstrate the performance of our system for garment transfer and pose-guided face resynthesis.

We present an end-to-end deep Convolutional Neural Network called Convolutional Relational Machine (CRM) for recognizing group activities that utilizes the information in spatial relations between individual persons in image or video. It learns to produce an intermediate spatial representation (activity map) based on individual and group activities. A multi-stage refinement component is responsible for decreasing the incorrect predictions in the activity map. Finally, an aggregation component uses the refined information to recognize group activities. Experimental results demonstrate the constructive contribution of the information extracted and represented in the form of the activity map. CRM shows advantages over state-of-the-art models on Volleyball and Collective Activity datasets.

Creating high definition maps that contain precise information of static elements of the scene is of utmost importance for enabling self driving cars to drive safely. In this paper, we tackle the problem of drivable road boundary extraction from LiDAR and camera imagery. Towards this goal, we design a structured model where a fully convolutional network obtains deep features encoding the location and direction of road boundaries and then, a convolutional recurrent network outputs a polyline representation for each one of them. Importantly, our method is fully automatic and does not require a user in the loop. We showcase the effectiveness of our method on a large North American city where we obtain perfect topology of road boundaries 99.3% of the time at a high precision and recall.

Visual illusions teach us that what we see is not always what is represented in the physical world. Their special nature make them a fascinating tool to test and validate any new vision model proposed. In general, current vision models are based on the concatenation of linear and non-linear operations. The similarity of this structure with the operations present in Convolutional Neural Networks (CNNs) has motivated us to study if CNNs trained for low-level visual tasks are deceived by visual illusions. In particular, we show that CNNs trained for image denoising, image deblurring, and computational color constancy are able to replicate the human response to visual illusions, and that the extent of this replication varies with respect to variation in architecture and spatial pattern size. These results suggest that in order to obtain CNNs that better replicate human behaviour, we may need to start aiming for them to better replicate visual illusions.

This paper addresses the problem of 3D human pose and shape estimation from a single image. Previous approaches consider a parametric model of the human body, SMPL, and attempt to regress the model parameters that give rise to a mesh consistent with image evidence. This parameter regression has been a very challenging task, with model-based approaches underperforming compared to nonparametric solutions in terms of pose estimation. In our work, we propose to relax this heavy reliance on the model's parameter space. We still retain the topology of the SMPL template mesh, but instead of predicting model parameters, we directly regress the 3D location of the mesh vertices. This is a heavy task for a typical network, but our key insight is that the regression becomes significantly easier using a Graph-CNN. This architecture allows us to explicitly encode the template mesh structure within the network and leverage the spatial locality the mesh has to offer. Image-based features are attached to the mesh vertices and the Graph-CNN is responsible to process them on the mesh structure, while the regression target for each vertex is its 3D location. Having recovered the complete 3D geometry of the mesh, if we still require a specific model parametrization, this can be reliably regressed from the vertices locations. We demonstrate the flexibility and the effectiveness of our proposed graph-based mesh regression by attaching different types of features on the mesh vertices. In all cases, we outperform the comparable baselines relying on model parameter regression, while we also achieve state-of-the-art results among model-based pose estimation approaches.

Unsupervised Domain Adaptation (UDA) makes predictions for the target domain data while manual annotations are only available in the source domain. Previous methods minimize the domain discrepancy neglecting the class information, which may lead to misalignment and poor generalization performance. To address this issue, this paper proposes Contrastive Adaptation Network (CAN) optimizing a new metric which explicitly models the intra-class domain discrepancy and the inter-class domain discrepancy. We design an alternating update strategy for training CAN in an end-to-end manner. Experiments on two real-world benchmarks Office-31 and VisDA-2017 demonstrate that CAN performs favorably against the state-of-the-art methods and produces more discriminative features.

The large availability of depth sensors provides valuable complementary information for salient object detection (SOD) in RGBD images. However, due to the inherent difference between RGB and depth information, extracting features from the depth channel using ImageNet pre-trained backbone models and fusing them with RGB features directly are sub-optimal. In this paper, we utilize contrast prior, which used to be a dominant cue in none deep learning based SOD approaches, into CNNs-based architecture to enhance the depth information. The enhanced depth cues are further integrated with RGB features for SOD, using a novel fluid pyramid integration, which can make better use of multi-scale cross-modal features. Comprehensive experiments on 5 challenging benchmark datasets demonstrate the superiority of the architecture CPFP over 9 state-of-the-art alternative methods.

Most existing studies on learning local features focus on the patch-based descriptions of individual keypoints, whereas neglecting the spatial relations established from their keypoint locations. In this paper, we go beyond the local detail representation by introducing context awareness to augment off-the-shelf local feature descriptors. Specifically, we propose a unified learning framework that leverages and aggregates the cross-modality contextual information, including (i) visual context from high-level image representation, and (ii) geometric context from 2D keypoint distribution. Moreover, we propose an effective N-pair loss that eschews the empirical hyper-parameter search and improves the convergence. The proposed augmentation scheme is lightweight compared with the raw local feature description, meanwhile improves remarkably on several large-scale benchmarks with diversified scenes, which demonstrates both strong practicality and generalization ability in geometric matching applications.

Recent efforts have shown the importance of context on deep convolutional neural network based semantic segmentation. Among others, the predicted segmentation map (p-map) itself which encodes rich high-level semantic cues (e.g. objects and layout) can be regarded as a promising source of context. In this paper, we propose a dedicated module, Context Net, to better explore the context information in p-maps. Without introducing any new supervisions, we formulate the context learning problem as a Markov Decision Process and optimize it using reinforcement learning during which the p-map and Context Net are treated as environment and agent, respectively. Through adequate explorations, the Context Net selects the information which has long-term benefit for segmentation inference. By incorporating the Context Net with a baseline segmentation scheme, we then propose a Context-reinforced Semantic Segmentation network (CiSS-Net), which is fully end-to-end trainable. Experimental results show that the learned context brings 3.9% absolute improvement on mIoU over the baseline segmentation method, and the CiSS-Net achieves the state-of-the-art segmentation performance on ADE20K, PASCAL-Context and Cityscapes.

How do we determine whether two or more clothing items are compatible or visually appealing? Part of the answer lies in understanding of visual aesthetics, and is biased by personal preferences shaped by social attitudes, time, and place. In this work we propose a method that predicts compatibility between two items based on their visual features, as well as their context. We define context as the products that are known to be compatible with each of these item. Our model is in contrast to other metric learning approaches that rely on pairwise comparisons between item features alone. We address the compatibility prediction problem using a graph neural network that learns to generate product embeddings conditioned on their context. We present results for two prediction tasks (fill in the blank and outfit compatibility) tested on two fashion datasets Polyvore and Fashion-Gen, and on a subset of the Amazon dataset; we achieve state of the art results when using context information and show how test performance improves as more context is used.

Curvilinear structures are frequently observed in various images in different forms, such as blood vessels or neuronal boundaries in biomedical images. In this paper, we propose a novel curvilinear structure segmentation approach using context-aware spatio-recurrent networks. Instead of directly segmenting the whole image or densely segmenting fixed-sized local patches, our method recurrently samples patches with varied scales from the target image with learned policy and processes them locally, which is similar to the behavior of changing retinal fixations in the human visual system and it is beneficial for capturing the multi-scale or hierarchical modality of the complex curvilinear structures. In specific, the policy of choosing local patches is attentively learned based on the contextual information of the image and the historical sampling experience. In this way, with more patches sampled and refined, the segmentation of the whole image can be progressively improved. To validate our approach, comparison experiments on different types of image data are conducted and the sampling procedures for exemplar images are illustrated. We demonstrate that our method achieves the state-of-the-art performance in public datasets.

State-of-the-art methods for counting people in crowded scenes rely on deep networks to estimate crowd density. They typically use the same filters over the whole image or over large image patches. Only then do they estimate local scale to compensate for perspective distortion. This is typically achieved by training an auxiliary classifier to select, for predefined image patches, the best kernel size among a limited set of choices. As such, these methods are not end-to-end trainable and restricted in the scope of context they can leverage. In this paper, we introduce an end-to-end trainable deep architecture that combines features obtained using multiple receptive field sizes and learns the importance of each such feature at each image location. In other words, our approach adaptively encodes the scale of the contextual information required to accurately predict crowd density. This yields an algorithm that outperforms state-of-the-art crowd counting methods, especially when perspective effects are strong.

Dense captioning aims at simultaneously localizing semantic regions and describing these regions-of-interest (ROIs) with short phrases or sentences in natural language. Previous studies have shown remarkable progresses, but they are often vulnerable to the aperture problem that a caption generated by the features inside one ROI lacks contextual coherence with its surrounding context in the input image. In this work, we investigate contextual reasoning based on multi-scale message propagations from the neighboring contents to the target ROIs. To this end, we design a novel end-to-end context and attribute grounded dense captioning framework consisting of 1) a contextual visual mining module and 2) a multi-level attribute grounded description generation module. Knowing that captions often co-occur with the linguistic attributes (such as who, what and where), we also incorporate an auxiliary supervision from hierarchical linguistic attributes to augment the distinctiveness of the learned captions. Extensive experiments and ablation studies on Visual Genome dataset demonstrate the superiority of the proposed model in comparison to state-of-the-art methods.