Showing papers by "Thomas Brox published in 2017"

FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks

Abstract: The FlowNet demonstrated that optical flow estimation can be cast as a learning problem. However, the state of the art with regard to the quality of the flow has still been defined by traditional methods. Particularly on small displacements and real-world data, FlowNet cannot compete with variational methods. In this paper, we advance the concept of end-to-end learning of optical flow and make it work really well. The large improvements in quality and speed are caused by three major contributions: first, we focus on the training data and show that the schedule of presenting data during training is very important. Second, we develop a stacked architecture that includes warping of the second image with intermediate optical flow. Third, we elaborate on small displacements by introducing a subnetwork specializing on small motions. FlowNet 2.0 is only marginally slower than the original FlowNet but decreases the estimation error by more than 50%. It performs on par with state-of-the-art methods, while running at interactive frame rates. Moreover, we present faster variants that allow optical flow computation at up to 140fps with accuracy matching the original FlowNet.

Octree Generating Networks: Efficient Convolutional Architectures for High-resolution 3D Outputs

Abstract: We present a deep convolutional decoder architecture that can generate volumetric 3D outputs in a compute- and memory-efficient manner by using an octree representation. The network learns to predict both the structure of the octree, and the occupancy values of individual cells. This makes it a particularly valuable technique for generating 3D shapes. In contrast to standard decoders acting on regular voxel grids, the architecture does not have cubic complexity. This allows representing much higher resolution outputs with a limited memory budget. We demonstrate this in several application domains, including 3D convolutional autoencoders, generation of objects and whole scenes from high-level representations, and shape from a single image.

DeMoN: Depth and Motion Network for Learning Monocular Stereo

Abstract: In this paper we formulate structure from motion as a learning problem. We train a convolutional network end-to-end to compute depth and camera motion from successive, unconstrained image pairs. The architecture is composed of multiple stacked encoder-decoder networks, the core part being an iterative network that is able to improve its own predictions. The network estimates not only depth and motion, but additionally surface normals, optical flow between the images and confidence of the matching. A crucial component of the approach is a training loss based on spatial relative differences. Compared to traditional two-frame structure from motion methods, results are more accurate and more robust. In contrast to the popular depth-from-single-image networks, DeMoN learns the concept of matching and, thus, better generalizes to structures not seen during training.

Learning to Estimate 3D Hand Pose from Single RGB Images

Abstract: Low-cost consumer depth cameras and deep learning have enabled reasonable 3D hand pose estimation from single depth images. In this paper, we present an approach that estimates 3D hand pose from regular RGB images. This task has far more ambiguities due to the missing depth information. To this end, we propose a deep network that learns a network-implicit 3D articulation prior. Together with detected keypoints in the images, this network yields good estimates of the 3D pose. We introduce a large scale 3D hand pose dataset based on synthetic hand models for training the involved networks. Experiments on a variety of test sets, including one on sign language recognition, demonstrate the feasibility of 3D hand pose estimation on single color images.

Sparsity Invariant CNNs

Abstract: In this paper, we consider convolutional neural networks operating on sparse inputs with an application to depth upsampling from sparse laser scan data. First, we show that traditional convolutional networks perform poorly when applied to sparse data even when the location of missing data is provided to the network. To overcome this problem, we propose a simple yet effective sparse convolution layer which explicitly considers the location of missing data during the convolution operation. We demonstrate the benefits of the proposed network architecture in synthetic and real experiments with respect to various baseline approaches. Compared to dense baselines, the proposed sparse convolution network generalizes well to novel datasets and is invariant to the level of sparsity in the data. For our evaluation, we derive a novel dataset from the KITTI benchmark, comprising 93k depth annotated RGB images. Our dataset allows for training and evaluating depth upsampling and depth prediction techniques in challenging real-world settings and will be made available upon publication.

An objective comparison of cell-tracking algorithms

Abstract: We present a combined report on the results of three editions of the Cell Tracking Challenge, an ongoing initiative aimed at promoting the development and objective evaluation of cell segmentation and tracking algorithms. With 21 participating algorithms and a data repository consisting of 13 data sets from various microscopy modalities, the challenge displays today's state-of-the-art methodology in the field. We analyzed the challenge results using performance measures for segmentation and tracking that rank all participating methods. We also analyzed the performance of all of the algorithms in terms of biological measures and practical usability. Although some methods scored high in all technical aspects, none obtained fully correct solutions. We found that methods that either take prior information into account using learning strategies or analyze cells in a global spatiotemporal video context performed better than other methods under the segmentation and tracking scenarios included in the challenge.

Octree Generating Networks: Efficient Convolutional Architectures for High-resolution 3D Outputs

Abstract: We present a deep convolutional decoder architecture that can generate volumetric 3D outputs in a compute- and memory-efficient manner by using an octree representation. The network learns to predict both the structure of the octree, and the occupancy values of individual cells. This makes it a particularly valuable technique for generating 3D shapes. In contrast to standard decoders acting on regular voxel grids, the architecture does not have cubic complexity. This allows representing much higher resolution outputs with a limited memory budget. We demonstrate this in several application domains, including 3D convolutional autoencoders, generation of objects and whole scenes from high-level representations, and shape from a single image.

Sparsity Invariant CNNs

Abstract: In this paper, we consider convolutional neural networks operating on sparse inputs with an application to depth upsampling from sparse laser scan data. First, we show that traditional convolutional networks perform poorly when applied to sparse data even when the location of missing data is provided to the network. To overcome this problem, we propose a simple yet effective sparse convolution layer which explicitly considers the location of missing data during the convolution operation. We demonstrate the benefits of the proposed network architecture in synthetic and real experiments with respect to various baseline approaches. Compared to dense baselines, the proposed sparse convolution network generalizes well to novel datasets and is invariant to the level of sparsity in the data. For our evaluation, we derive a novel dataset from the KITTI benchmark, comprising 93k depth annotated RGB images. Our dataset allows for training and evaluating depth upsampling and depth prediction techniques in challenging real-world settings and will be made available upon publication.

Chained Multi-stream Networks Exploiting Pose, Motion, and Appearance for Action Classification and Detection

Abstract: General human action recognition requires understanding of various visual cues. In this paper, we propose a network architecture that computes and integrates the most important visual cues for action recognition: pose, motion, and the raw images. For the integration, we introduce a Markov chain model which adds cues successively. The resulting approach is efficient and applicable to action classification as well as to spatial and temporal action localization. The two contributions clearly improve the performance over respective baselines. The overall approach achieves state-of-the-art action classification performance on HMDB51, J-HMDB and NTU RGB+D datasets. Moreover, it yields state-of-the-art spatio-temporal action localization results on UCF101 and J-HMDB.

Universal Adversarial Perturbations Against Semantic Image Segmentation

Abstract: While deep learning is remarkably successful on perceptual tasks, it was also shown to be vulnerable to adversarial perturbations of the input. These perturbations denote noise added to the input that was generated specifically to fool the system while being quasi-imperceptible for humans. More severely, there even exist universal perturbations that are input-agnostic but fool the network on the majority of inputs. While recent work has focused on image classification, this work proposes attacks against semantic image segmentation: we present an approach for generating (universal) adversarial perturbations that make the network yield a desired target segmentation as output. We show empirically that there exist barely perceptible universal noise patterns which result in nearly the same predicted segmentation for arbitrary inputs. Furthermore, we also show the existence of universal noise which removes a target class (e.g., all pedestrians) from the segmentation while leaving the segmentation mostly unchanged otherwise.

Learning to Generate Chairs, Tables and Cars with Convolutional Networks

Abstract: We train generative ‘up-convolutional’ neural networks which are able to generate images of objects given object style, viewpoint, and color. We train the networks on rendered 3D models of chairs, tables, and cars. Our experiments show that the networks do not merely learn all images by heart, but rather find a meaningful representation of 3D models allowing them to assess the similarity of different models, interpolate between given views to generate the missing ones, extrapolate views, and invent new objects not present in the training set by recombining training instances, or even two different object classes. Moreover, we show that such generative networks can be used to find correspondences between different objects from the dataset, outperforming existing approaches on this task.

Universal Adversarial Perturbations Against Semantic Image Segmentation

Abstract: While deep learning is remarkably successful on perceptual tasks, it was also shown to be vulnerable to adversarial perturbations of the input. These perturbations denote noise added to the input that was generated specifically to fool the system while being quasi-imperceptible for humans. More severely, there even exist universal perturbations that are input-agnostic but fool the network on the majority of inputs. While recent work has focused on image classification, this work proposes attacks against semantic image segmentation: we present an approach for generating (universal) adversarial perturbations that make the network yield a desired target segmentation as output. We show empirically that there exist barely perceptible universal noise patterns which result in nearly the same predicted segmentation for arbitrary inputs. Furthermore, we also show the existence of universal noise which removes a target class (e.g., all pedestrians) from the segmentation while leaving the segmentation mostly unchanged otherwise.

Orientation-boosted Voxel Nets for 3D Object Recognition

Abstract: Recent work has shown good recognition results in 3D object recognition using 3D convolutional networks. In this paper, we show that the object orientation plays an important role in 3D recognition. More specifically, we argue that objects induce different features in the network under rotation. Thus, we approach the category-level classification task as a multi-task problem, in which the network is trained to predict the pose of the object in addition to the class label as a parallel task. We show that this yields significant improvements in the classification results. We test our suggested architecture on several datasets representing various 3D data sources: LiDAR data, CAD models, and RGB-D images. We report state-of-the-art results on classification as well as significant improvements in precision and speed over the baseline on 3D detection.

Chained Multi-stream Networks Exploiting Pose, Motion, and Appearance for Action Classification and Detection

Abstract: General human action recognition requires understanding of various visual cues. In this paper, we propose a network architecture that computes and integrates the most important visual cues for action recognition: pose, motion, and the raw images. For the integration, we introduce a Markov chain model which adds cues successively. The resulting approach is efficient and applicable to action classification as well as to spatial and temporal action localization. The two contributions clearly improve the performance over respective baselines. The overall approach achieves state-of-the-art action classification performance on HMDB51, J-HMDB and NTU RGB+D datasets. Moreover, it yields state-of-the-art spatio-temporal action localization results on UCF101 and J-HMDB.

Semantics-aware visual localization under challenging perceptual conditions

Abstract: Visual place recognition under difficult perceptual conditions remains a challenging problem due to changing weather conditions, illumination and seasons Long-term visual navigation approaches for robot localization should be robust to these dynamics of the environment Existing methods typically leverage feature descriptions of whole images or image regions from Deep Convolutional Neural Networks Some approaches also exploit sequential information to alleviate the problem of spatially inconsistent and non-perfect image matches In this paper, we propose a novel approach for learning a discriminative holistic image representation which exploits the image content to create a dense and salient scene description These salient descriptions are learnt over a variety of datasets under large perceptual changes Such an approach enables us to precisely segment the regions of an image which are geometrically stable over large time lags We combine features from these salient regions and an off-the-shelf holistic representation to form a more robust scene descriptor We also introduce a semantically labeled dataset which captures extreme perceptual and structural scene dynamics over the course of 3 years We evaluated our approach with extensive experiments on data collected over several kilometers in Freiburg and show that our learnt image representation outperforms off-the-shelf features from the deep networks and hand-crafted features

Lucid Data Dreaming for Object Tracking

Abstract: Convolutional networks reach top quality in pixel-level video object segmentation but require a large amount of training data (1k~100k) to deliver such results. We propose a new training strategy which achieves state-of-the-art results across three evaluation datasets while using 20x~1000x less annotated data than competing methods. Our approach is suitable for both single and multiple object segmentation. Instead of using large training sets hoping to generalize across domains, we generate in-domain training data using the provided annotation on the first frame of each video to synthesize ("lucid dream") plausible future video frames. In-domain per-video training data allows us to train high quality appearance- and motion-based models, as well as tune the post-processing stage. This approach allows to reach competitive results even when training from only a single annotated frame, without ImageNet pre-training. Our results indicate that using a larger training set is not automatically better, and that for the video object segmentation task a smaller training set that is closer to the target domain is more effective. This changes the mindset regarding how many training samples and general "objectness" knowledge are required for the video object segmentation task.

Joint Graph Decomposition & Node Labeling: Problem, Algorithms, Applications

Abstract: We state a combinatorial optimization problem whose feasible solutions define both a decomposition and a node labeling of a given graph. This problem offers a common mathematical abstraction of seemingly unrelated computer vision tasks, including instance-separating semantic segmentation, articulated human body pose estimation and multiple object tracking. Conceptually, it generalizes the unconstrained integer quadratic program and the minimum cost lifted multicut problem, both of which are NP-hard. In order to find feasible solutions efficiently, we define two local search algorithms that converge monotonously to a local optimum, offering a feasible solution at any time. To demonstrate the effectiveness of these algorithms in tackling computer vision tasks, we apply them to instances of the problem that we construct from published data, using published algorithms. We report state-of-the-art application-specific accuracy in the three above-mentioned applications.

Lucid Data Dreaming for Object Tracking

Abstract: Convolutional networks reach top quality in pixel-level video object segmentation but require a large amount of training data (1k~100k) to deliver such results. We propose a new training strategy which achieves state-of-the-art results across three evaluation datasets while using 20x~1000x less annotated data than competing methods. Our approach is suitable for both single and multiple object segmentation. Instead of using large training sets hoping to generalize across domains, we generate in-domain training data using the provided annotation on the first frame of each video to synthesize ("lucid dream") plausible future video frames. In-domain per-video training data allows us to train high quality appearance- and motion-based models, as well as tune the post-processing stage. This approach allows to reach competitive results even when training from only a single annotated frame, without ImageNet pre-training. Our results indicate that using a larger training set is not automatically better, and that for the video object segmentation task a smaller training set that is closer to the target domain is more effective. This changes the mindset regarding how many training samples and general "objectness" knowledge are required for the video object segmentation task.

Lucid Data Dreaming for Video Object Segmentation

Abstract: Convolutional networks reach top quality in pixel-level video object segmentation but require a large amount of training data (1k~100k) to deliver such results. We propose a new training strategy which achieves state-of-the-art results across three evaluation datasets while using 20x~1000x less annotated data than competing methods. Our approach is suitable for both single and multiple object segmentation. Instead of using large training sets hoping to generalize across domains, we generate in-domain training data using the provided annotation on the first frame of each video to synthesize ("lucid dream") plausible future video frames. In-domain per-video training data allows us to train high quality appearance- and motion-based models, as well as tune the post-processing stage. This approach allows to reach competitive results even when training from only a single annotated frame, without ImageNet pre-training. Our results indicate that using a larger training set is not automatically better, and that for the video object segmentation task a smaller training set that is closer to the target domain is more effective. This changes the mindset regarding how many training samples and general "objectness" knowledge are required for the video object segmentation task.

Adversarial Examples for Semantic Image Segmentation

Abstract: Machine learning methods in general and Deep Neural Networks in particular have shown to be vulnerable to adversarial perturbations. So far this phenomenon has mainly been studied in the context of whole-image classification. In this contribution, we analyse how adversarial perturbations can affect the task of semantic segmentation. We show how existing adversarial attackers can be transferred to this task and that it is possible to create imperceptible adversarial perturbations that lead a deep network to misclassify almost all pixels of a chosen class while leaving network prediction nearly unchanged outside this class.

Learning to Estimate 3D Hand Pose from Single RGB Images

Abstract: Low-cost consumer depth cameras and deep learning have enabled reasonable 3D hand pose estimation from single depth images. In this paper, we present an approach that estimates 3D hand pose from regular RGB images. This task has far more ambiguities due to the missing depth information. To this end, we propose a deep network that learns a network-implicit 3D articulation prior. Together with detected keypoints in the images, this network yields good estimates of the 3D pose. We introduce a large scale 3D hand pose dataset based on synthetic hand models for training the involved networks. Experiments on a variety of test sets, including one on sign language recognition, demonstrate the feasibility of 3D hand pose estimation on single color images.

Global, Dense Multiscale Reconstruction for a Billion Points

Abstract: We present a variational approach for surface reconstruction from a set of oriented points with scale information. We focus particularly on scenarios with nonuniform point densities due to images taken from different distances. In contrast to previous methods, we integrate the scale information in the objective and globally optimize the signed distance function of the surface on a balanced octree grid. We use a finite element discretization on the dual structure of the octree minimizing the number of variables. The tetrahedral mesh is generated efficiently with a lookup table which allows to map octree cells to the nodes of the finite elements. We optimize memory efficiency by data aggregation, such that robust data terms can be used even on very large scenes. The surface normals are explicitly optimized and used for surface extraction to improve the reconstruction at edges and corners.

End-to-End Learning of Video Super-Resolution with Motion Compensation

Abstract: Learning approaches have shown great success in the task of super-resolving an image given a low resolution input. Video super-resolution aims for exploiting additionally the information from multiple images. Typically, the images are related via optical flow and consecutive image warping. In this paper, we provide an end-to-end video super-resolution network that, in contrast to previous works, includes the estimation of optical flow in the overall network architecture. We analyze the usage of optical flow for video super-resolution and find that common off-the-shelf image warping does not allow video super-resolution to benefit much from optical flow. We rather propose an operation for motion compensation that performs warping from low to high resolution directly. We show that with this network configuration, video super-resolution can benefit from optical flow and we obtain state-of-the-art results on the popular test sets. We also show that the processing of whole images rather than independent patches is responsible for a large increase in accuracy.

End-to-End Learning of Video Super-Resolution with Motion Compensation

Abstract: Learning approaches have shown great success in the task of super-resolving an image given a low resolution input. Video super-resolution aims for exploiting additionally the information from multiple images. Typically, the images are related via optical flow and consecutive image warping. In this paper, we provide an end-to-end video super-resolution network that, in contrast to previous works, includes the estimation of optical flow in the overall network architecture. We analyze the usage of optical flow for video super-resolution and find that common off-the-shelf image warping does not allow video super-resolution to benefit much from optical flow. We rather propose an operation for motion compensation that performs warping from low to high resolution directly. We show that with this network configuration, video super-resolution can benefit from optical flow and we obtain state-of-the-art results on the popular test sets. We also show that the processing of whole images rather than independent patches is responsible for a large increase in accuracy.

Lucid Data Dreaming for Multiple Object Tracking

Abstract: Convolutional networks reach top quality in pixel-level video object segmentation but require a large amount of training data (1k~100k) to deliver such results. We propose a new training strategy which achieves state-of-the-art results across three evaluation datasets while using 20x~1000x less annotated data than competing methods. Our approach is suitable for both single and multiple object segmentation. Instead of using large training sets hoping to generalize across domains, we generate in-domain training data using the provided annotation on the first frame of each video to synthesize ("lucid dream") plausible future video frames. In-domain per-video training data allows us to train high quality appearance- and motion-based models, as well as tune the post-processing stage. This approach allows to reach competitive results even when training from only a single annotated frame, without ImageNet pre-training. Our results indicate that using a larger training set is not automatically better, and that for the video object segmentation task a smaller training set that is closer to the target domain is more effective. This changes the mindset regarding how many training samples and general "objectness" knowledge are required for the video object segmentation task.

Topometric Localization with Deep Learning

Abstract: Compared to LiDAR-based localization methods, which provide high accuracy but rely on expensive sensors, visual localization approaches only require a camera and thus are more cost-effective however their accuracy and reliability is typically inferior to LiDAR-based methods. In this work, we propose a vision-based localization approach that learns from LiDAR-based localization methods by using their output as training data, thus combining a cheap, passive sensor with an accuracy that is on-par with LiDAR-based localization. The approach consists of two deep networks trained on visual odometry and topological localization, respectively, and a successive optimization to combine the predictions of these two networks. Furthermore, we introduce a new challenging pedestrian-based dataset for localization with a high degree of noise. Results obtained by evaluating the proposed approach on this novel dataset demonstrate localization errors up to 10 times smaller than those obtained with traditional vision-based localization methods.

Adversarial Examples for Semantic Image Segmentation

Abstract: Machine learning methods in general and Deep Neural Networks in particular have shown to be vulnerable to adversarial perturbations. So far this phenomenon has mainly been studied in the context of whole-image classification. In this contribution, we analyse how adversarial perturbations can affect the task of semantic segmentation. We show how existing adversarial attackers can be transferred to this task and that it is possible to create imperceptible adversarial perturbations that lead a deep network to misclassify almost all pixels of a chosen class while leaving network prediction nearly unchanged outside this class.

Non-smooth Non-convex Bregman Minimization: Unification and new Algorithms

Abstract: We propose a unifying algorithm for non-smooth non-convex optimization. The algorithm approximates the objective function by a convex model function and finds an approximate (Bregman) proximal point of the convex model. This approximate minimizer of the model function yields a descent direction, along which the next iterate is found. Complemented with an Armijo-like line search strategy, we obtain a flexible algorithm for which we prove (subsequential) convergence to a stationary point under weak assumptions on the growth of the model function error. Special instances of the algorithm with a Euclidean distance function are, for example, Gradient Descent, Forward--Backward Splitting, ProxDescent, without the common requirement of a "Lipschitz continuous gradient". In addition, we consider a broad class of Bregman distance functions (generated by Legendre functions) replacing the Euclidean distance. The algorithm has a wide range of applications including many linear and non-linear inverse problems in signal/image processing and machine learning.

Artistic style transfer for videos and spherical images

Abstract: Manually re-drawing an image in a certain artistic style takes a professional artist a long time. Doing this for a video sequence single-handedly is beyond imagination. We present two computational approaches that transfer the style from one image (for example, a painting) to a whole video sequence. In our first approach, we adapt to videos the original image style transfer technique by Gatys et al. based on energy minimization. We introduce new ways of initialization and new loss functions to generate consistent and stable stylized video sequences even in cases with large motion and strong occlusion. Our second approach formulates video stylization as a learning problem. We propose a deep network architecture and training procedures that allow us to stylize arbitrary-length videos in a consistent and stable way, and nearly in real time. We show that the proposed methods clearly outperform simpler baselines both qualitatively and quantitatively. Finally, we propose a way to adapt these approaches also to 360 degree images and videos as they emerge with recent virtual reality hardware.

Topometric Localization with Deep Learning

Abstract: Compared to LiDAR-based localization methods, which provide high accuracy but rely on expensive sensors, visual localization approaches only require a camera and thus are more cost-effective while their accuracy and reliability typically is inferior to LiDAR-based methods. In this work, we propose a vision-based localization approach that learns from LiDAR-based localization methods by using their output as training data, thus combining a cheap, passive sensor with an accuracy that is on-par with LiDAR-based localization. The approach consists of two deep networks trained on visual odometry and topological localization, respectively, and a successive optimization to combine the predictions of these two networks. We evaluate the approach on a new challenging pedestrian-based dataset captured over the course of six months in varying weather conditions with a high degree of noise. The experiments demonstrate that the localization errors are up to 10 times smaller than with traditional vision-based localization methods.