CVPR 2017 Papers

We introduce a neural architecture for navigation in novel environments. Our proposed architecture learns to map from first-person views and plans a sequence of actions towards goals in the environment. The Cognitive Mapper and Planner (CMP) is based on two key ideas: a) a unified joint architecture for mapping and planning, such that the mapping is driven by the needs of the planner, and b) a spatial memory with the ability to plan given an incomplete set of observations about the world. CMP constructs a top-down belief map of the world and applies a differentiable neural net planner to produce the next action at each time step. The accumulated belief of the world enables the agent to track visited regions of the environment. Our experiments demonstrate that CMP outperforms both reactive strategies and standard memory-based architectures and performs well in novel environments. Furthermore, we show that CMP can also achieve semantically specified goals, such as "go to a chair".

This paper addresses the challenge of 3D human pose estimation from a single color image. Despite the general success of the end-to-end learning paradigm, top performing approaches employ a two-step solution consisting of a Convolutional Network (ConvNet) for 2D joint localization and a subsequent optimization step to recover 3D pose. In this paper, we identify the representation of 3D pose as a critical issue with current ConvNet approaches and make two important contributions towards validating the value of end-to-end learning for this task. First, we propose a fine discretization of the 3D space around the subject and train a ConvNet to predict per voxel likelihoods for each joint. This creates a natural representation for 3D pose and greatly improves performance over the direct regression of joint coordinates. Second, to further improve upon initial estimates, we employ a coarse-to-fine prediction scheme. This step addresses the large dimensionality increase and enables iterative refinement and repeated processing of the image features. The proposed approach outperforms all state-of-the-art methods on standard benchmarks achieving a relative error reduction greater than 30% on average. Additionally, we investigate using our volumetric representation in a related architecture which is suboptimal compared to our end-to-end approach, but is of practical interest, since it enables training when no image with corresponding 3D groundtruth is available, and allows us to present compelling results for in-the-wild images.

We formulate an energy for segmentation that is designed to have preference for segmenting the coarse over fine structure of the image, without smoothing across boundaries of regions. The energy is formulated by integrating a continuum of scales from a scale space computed from the heat equation within regions. We show that the energy can be optimized without computing a continuum of scales, but instead from a single scale. This makes the method computationally efficient in comparison to energies using a discrete set of scales. We apply our method to texture and motion segmentation. Experiments on benchmark datasets show that a continuum of scales leads to better segmentation accuracy over discrete scales and other competing methods.

Co-occurrence Filter (CoF) is a boundary preserving filter. It is based on the Bilateral Filter (BF) but instead of using a Gaussian on the range values to preserve edges it relies on a co-occurrence matrix. Pixel values that co-occur frequently in the image (i.e., inside textured regions) will have a high weight in the co-occurrence matrix. This, in turn, means that such pixel pairs will be averaged and hence smoothed, regardless of their intensity differences. On the other hand, pixel values that rarely co-occur (i.e., across texture boundaries) will have a low weight in the co-occurrence matrix. As a result, they will not be averaged and the boundary between them will be preserved. The CoF therefore extends the BF to deal with boundaries, not just edges. It learns co-occurrences directly from the image. We can achieve various filtering results by directing it to learn the co-occurrence matrix from a part of the image, or a different image. We give the definition of the filter, discuss how to use it with color images and show several use cases.

Convnets have enabled significant progress in pedestrian detection recently, but there are still open questions regard- ing suitable architectures and training data. We revisit CNN design and point out key adaptations, enabling plain Fas- terRCNN to obtain state-of-the-art results on the Caltech dataset. To achieve further improvement from more and better data, we introduce CityPersons, a new set of person annotations on top of the Cityscapes dataset. The di- versity of CityPersons allows us for the first time to train one single CNN model that generalizes well over mul- tiple benchmarks. Moreover, with additional training with CityPersons, we obtain top results using FasterRCNN on Caltech, improving especially for more difficult cases (heavy occlusion and small scale) and providing higher loc- alization quality.

The chest X-ray is one of the most commonly accessible radiological examinations for screening and diagnosis of many lung diseases. A tremendous number of X-ray imaging studies accompanied by radiological reports are accumulated and stored in many modern hospitals' Picture Archiving and Communication Systems (PACS). On the other side, it is still an open question how this type of hospital-size knowledge database containing invaluable imaging informatics (i.e., loosely labeled) can be used to facilitate the data-hungry deep learning paradigms in building truly large-scale high precision computer-aided diagnosis (CAD) systems. In this paper, we present a new chest X-ray database, namely "ChestX-ray8", which comprises 108,948 frontal-view X-ray images of 32,717 unique patients with the text-mined eight disease image labels (where each image can have multi-labels), from the associated radiological reports using natural language processing. Importantly, we demonstrate that these commonly occurring thoracic diseases can be detected and even spatially-located via a unified weakly-supervised multi-label image classification and disease localization framework, which is validated using our proposed dataset. Although the initial quantitative results are promising as reported, deep convolutional neural network based "reading chest X-rays" (i.e., recognizing and locating the common disease patterns trained with only image-level labels) remains a strenuous task for fully-automated high precision CAD systems.

Recent captioning models are limited in their ability to scale and describe concepts unseen in paired image-text corpora. We propose the Novel Object Captioner (NOC), a deep visual semantic captioning model that can describe a large number of object categories not present in existing image-caption datasets. Our model takes advantage of external sources -- labeled images from object recognition datasets, and semantic knowledge extracted from unannotated text. We propose minimizing a joint objective which can learn from these diverse data sources and leverage distributional semantic embeddings, enabling the model to generalize and describe novel objects outside of image-caption datasets. We demonstrate that our model exploits semantic information to generate captions for hundreds of object categories in the ImageNet object recognition dataset that are not observed in MSCOCO image-caption training data, as well as many categories that are observed very rarely. Both automatic evaluations and human judgements show that our model considerably outperforms prior work in being able to describe many more categories of objects.

Effectively describing and recognizing leaf shapes under arbitrary deformations, particularly from a large database, remains an unsolved problem. In this research, we attempted a new strategy of describing shape by walking along a bunch of chords that pass through the shape to measure the regions trespassed. A novel chord bunch walks (CBW) descriptor is developed through the chord walking that effectively integrates the shape image function over the walked chord to reflect the contour features and the inner properties of the shape. For each contour point, the chord bunch groups multiple pairs of chord walks to build a hierarchical framework for a coarse-to-fine description. The proposed CBW descriptor is invariant to rotation, scaling, translation, and mirror transforms. Instead of using the expensive optimal correspondence based matching, an improved Hausdorff distance encoded correspondence information is proposed for efficient yet effective shape matching. In experimental studies, the proposed method obtained substantially higher accuracies with low computational cost over the benchmarks, which indicates the research potential along this direction.

Given the recent advances in depth prediction from Convolutional Neural Networks (CNNs), this paper investigates how predicted depth maps from a deep neural network can be deployed for the goal of accurate and dense monocular reconstruction. We propose a method where CNN-predicted dense depth maps are naturally fused together with depth measurements obtained from direct monocular SLAM, based on a scheme that privileges depth prediction in image locations where monocular SLAM approaches tend to fail, e.g. along low-textured regions, and vice-versa. We demonstrate the use of depth prediction to estimate the absolute scale of the reconstruction, hence overcoming one of the major limitations of monocular SLAM. Finally, we propose a framework to efficiently fuse semantic labels, obtained from a single frame, with dense SLAM, so to yield semantically coherent scene reconstruction from a single view. Evaluation results on two benchmark datasets show the robustness and accuracy of our approach.

Learning based approaches have not yet achieved their full potential in optical flow estimation, where their performance still trails heuristic approaches. In this paper, we present a CNN based patch matching approach for optical flow estimation. An important contribution of our approach is a novel thresholded loss for Siamese networks. We demonstrate that our loss performs clearly better than existing losses. It also allows to speed up training by a factor of 2 in our tests. Furthermore, we present a novel way for calculating CNN based features for different image scales, which performs better than existing methods. We also discuss new ways of evaluating the robustness of trained features for the application of patch matching for optical flow. An interesting discovery in our paper is that low-pass filtering of feature maps can increase the robustness of features created by CNNs. We proved the competitive performance of our approach by submitting it to the KITTI 2012, KITTI 2015 and MPI-Sintel evaluation portals where we obtained state-of-the-art results on all three datasets.

This paper proposes a data-driven approach for image alignment. Our main contribution is a novel network architecture that combines the strengths of convolutional neural networks (CNNs) and the Lucas-Kanade algorithm. The main component of this architecture is a Lucas-Kanade layer that performs the inverse compositional algorithm on convolutional feature maps. To train our network, we develop a cascaded feature learning method that incorporates the coarse-to-fine strategy into the training process. This method learns a pyramid representation of convolutional features in a cascaded manner and yields a cascaded network that performs coarse-to-fine alignment on the feature pyramids. We apply our model to the task of homography estimation, and perform training and evaluation on a large labeled dataset generated from the MS-COCO dataset. Experimental results show that the proposed approach significantly outperforms the other methods.

When building artificial intelligence systems that can reason and answer questions about visual data, we need diagnostic tests to analyze our progress and discover short- comings. Existing benchmarks for visual question answer- ing can help, but have strong biases that models can exploit to correctly answer questions without reasoning. They also conflate multiple sources of error, making it hard to pinpoint model weaknesses. We present a diagnostic dataset that tests a range of visual reasoning abilities. It contains minimal biases and has detailed annotations describing the kind of reasoning each question requires. We use this dataset to analyze a variety of modern visual reasoning systems, providing novel insights into their abilities and limitations.

This work is about recognizing human activities occurring in videos at distinct semantic levels, including individual actions, interactions, and group activities. The recognition is realized using a two-level hierarchy of Long Short-Term Memory (LSTM) networks, forming a feed-forward deep architecture, which can be trained end-to-end. In comparison with existing architectures of LSTMs, we make two key contributions giving the name to our approach as Confidence-Energy Recurrent Network -- CERN. First, instead of using the common softmax layer for prediction, we specify a novel energy layer (EL) for estimating the energy of our predictions. Second, rather than finding the common minimum-energy class assignment, which may be numerically unstable under uncertainty, we specify that the EL additionally computes the p-values of the solutions, and in this way estimates the most confident energy minimum. The evaluation on the Collective Activity and Volleyball datasets demonstrates: (i) advantages of our two contributions relative to the common softmax and energy-minimization formulations and (ii) a superior performance relative to the state-of-the-art approaches.

Temporal action localization is an important yet challenging problem. Given a long, untrimmed video consisting of multiple action instances and complex background contents, we need not only to recognize their action categories, but also to localize the start time and end time of each instance. Many state-of-the-art systems use segment-level classifiers to select and rank proposal segments of pre-determined boundaries. However, a desirable model should move beyond segment-level and make dense predictions at a fine granularity in time to determine precise temporal boundaries. To this end, we design a novel Convolutional-De-Convolutional (CDC) network that places CDC filters on top of 3D ConvNets, which have been shown to be effective for abstracting action semantics but reduce the temporal length of the input data. The proposed CDC filter performs the required temporal upsampling and spatial downsampling operations simultaneously to predict actions at the frame-level granularity. It is unique in jointly modeling action semantics in space-time and fine-grained temporal dynamics. We train the CDC network in an end-to-end manner efficiently. Our model not only achieves superior performance in detecting actions in every frame, but also significantly boosts the precision of localizing temporal boundaries. Finally, the CDC network demonstrates a very high efficiency with the ability to process 500 frames per second on a single GPU server. Source code and trained models are available online at https://bitbucket.org/columbiadvmm/cdc.

Stereo matching is a well researched area using visible-band color cameras. Thermal images are typically lower resolution, have less texture, and are noisier compared to their visible-band counterparts and are more challenging for stereo matching algorithms. Previous benchmarks for stereo matching either focus entirely on visible-band cameras or contain only a single thermal camera. We present the Color And Thermal Stereo (CATS) benchmark, a dataset consisting of stereo thermal, stereo color, and cross-modality image pairs with high accuracy ground truth (< 2mm) generated from a LiDAR. We scanned 100 cluttered indoor and 80 outdoor scenes featuring challenging environments and conditions. CATS contains approximately 1400 images of pedestrians, vehicles, electronics, and other thermally interesting objects in different environmental conditions, including nighttime, daytime, and foggy scenes. Ground truth was projected to each of the four cameras to generate color-color, thermal-thermal, and cross-modality disparity maps. We develop a semi-automatic LiDAR to camera alignment procedure that does not require a calibration target. We compare state-of-the-art algorithms to baseline the dataset and show that in the thermal and cross modalities there is still much room for improvement. We expect our dataset to provide researchers with a more diverse set of imaged locations, objects, and modalities than previous benchmarks for stereo matching.

Boundary and edge cues are highly beneficial in improving a wide variety of vision tasks such as semantic segmentation, object recognition, stereo, and object proposal generation. Recently, the problem of edge detection has been revisited and significant progress has been made with deep learning. While classical edge detection is a challenging binary problem in itself, the category-aware semantic edge detection by nature is an even more challenging multi-label problem. We model the problem such that each edge pixel can be associated with more than one class as they appear in contours or junctions belonging to two or more semantic classes. To this end, we propose a novel end-to-end deep semantic edge learning architecture based on ResNet and a new skip-layer architecture where category-wise edge activations at the top convolution layer share and are fused with the same set of bottom layer features. We then propose a multi-label loss function to supervise the fused activations. We show that our proposed architecture benefits this problem with better performance, and we outperform the current state-of-the-art semantic edge detection methods by a large margin on standard data sets such as SBD and Cityscapes.

In this work, we build a generic architecture of Convolutional Neural Networks to discover empirical properties of neural networks. Our first contribution is to introduce a state-of-the-art framework that depends upon few hyper parameters and to study the network when we vary them. It has no max pooling, no biases, only 13 layers, is purely convolutional and yields up to 95.4% and 79.6% accuracy respectively on CIFAR10 and CIFAR100. We show that the nonlinearity of a deep network does not need to be continuous, non expansive or point-wise, to achieve good performance. We show that increasing the width of our network permits being competitive with very deep networks. Our second contribution is an analysis of the contraction and separation properties of this network. Indeed, a 1-nearest neighbor classifier applied on deep features progressively improves with depth, which indicates that the representation is progressively more regular. Besides, we defined and analyzed local support vectors that separate classes locally. All our experiments are reproducible and code will be available online, based on TensorFlow.

In this work, we study a poorly understood trade-off between accuracy and runtime costs for deep semantic video segmentation. While recent work has demonstrated advantages of learning to speed-up deep activity detection, it is not clear if similar advantages will hold for our very different segmentation loss function, which is defined over individual pixels across the frames. In deep video segmentation, the most time consuming step represents the application of a CNN to every frame for assigning class labels to every pixel, typically taking 6-9 times of the video footage. This motivates our new budget-aware framework that learns to optimally select a small subset of frames for pixelwise labeling by a CNN, and then efficiently interpolates the obtained segmentations to yet unprocessed frames. This interpolation may use either a simple optical-flow guided mapping of pixel labels, or another significantly less complex and thus faster CNN. We formalize the frame selection as a Markov Decision Process, and specify a Long Short-Term Memory (LSTM) network to model a policy for selecting the frames. For training the LSTM, we develop a policy-gradient reinforcement-learning approach for approximating the gradient of our non-decomposable and non-differentiable objective. Evaluation on two benchmark video datasets show that our new framework is able to significantly reduce computation time, and maintain competitive video segmentation accuracy under varying budgets.

We propose an extremely simple but effective regularization technique of convolutional neural networks (CNNs), referred to as BranchOut, for online ensemble tracking. Our algorithm employs a CNN for target representation, which has a common convolutional layers but has multiple branches of fully connected layers. For better regularization, a subset of branches in the CNN are selected randomly for online learning whenever target appearance models need to be updated. Each branch may have a different number of layers to maintain variable abstraction levels of target appearances. BranchOut with multi-level target representation allows us to learn robust target appearance models with diversity and handle various challenges in visual tracking problem effectively. The proposed algorithm is evaluated in standard tracking benchmarks and shows the state-of-the-art performance even without additional pretraining on external tracking sequences.