Search

The overarching goals in image-based localization are scale, robustness, and speed. In recent years, approaches based on local features and sparse 3D point-cloud models have both dominated the benchmarks and seen successful real-world deployment. They enable applications ranging from robot navigation, autonomous driving, virtual and augmented reality to device geo-localization. Recently, end-to-end learned localization approaches have been proposed which show promising results on small-scale datasets. However, the positioning accuracy, scalability, latency, and compute and storage requirements of these approaches remain open challenges. We aim to deploy localization at a global scale where one thus relies on methods using local features and sparse 3D models. Our approach spans from offline model building to real-time client-side pose fusion. The system compresses the appearance and geometry of the scene for efficient model storage and lookup leading to scalability beyond what has been demonstrated previously. It allows for low-latency localization queries and efficient fusion to be run in real-time on mobile platforms by combining server-side localization with real-time visual–inertial-based camera pose tracking. In order to further improve efficiency, we leverage a combination of priors, nearest-neighbor search, geometric match culling, and a cascaded pose candidate refinement step. This combination outperforms previous approaches when working with large-scale models and allows deployment at unprecedented scale. We demonstrate the effectiveness of our approach on a proof-of-concept system localizing 2.5 million images against models from four cities in different regions of the world achieving query latencies in the 200 ms range.

Camera pose estimation in known scenes is a 3D geometry task recently tackled by multiple learning algorithms. Many regress precise geometric quantities, like poses or 3D points, from an input image. This either fails to generalize to new viewpoints or ties the model parameters to a specific scene. In this paper, we go Back to the Feature: we argue that deep networks should focus on learning robust and invariant visual features, while the geometric estimation should be left to principled algorithms. We introduce PixLoc, a scene-agnostic neural network that estimates an accurate 6-DoF pose from an image and a 3D model. Our approach is based on the direct alignment of multiscale deep features, casting camera localization as metric learning. PixLoc learns strong data priors by end-to-end training from pixels to pose and exhibits exceptional generalization to new scenes by separating model parameters and scene geometry. The system can localize in large environments given coarse pose priors but also improve the accuracy of sparse feature matching by jointly refining keypoints and poses with little overhead. The code will be publicly available at https://github.com/cvg/pixloc., Comment: Accepted to CVPR 2021

Decomposing scenes into reflectance and lighting is an important task for applications such as relighting, image matching or content creation. Advanced light transport effects like occlusion and indirect lighting are often ignored, leading to subpar decompositions in which the albedo needs to compensate for insufficiencies in the estimated shading. We show how to account for these advanced lighting effects by utilizing precomputed radiance transfer to estimate reflectance and lighting. Given the geometry of an object and one or multiple images, our method reconstructs the object’s surface reflectance properties—such as its albedo and glossiness—as well as a colored lighting environment map. Evaluation on synthetic and real data shows that incorporation of indirect light leads to qualitatively and quantitatively improved results.

Surface reconstruction from point clouds has been a well-studied research topic with applications in computer vision and computer graphics. Recently, several learningbased methods were proposed for 3D shape representation through implicit functions which among others can be used for point cloud-based reconstruction. Although delivering compelling results for synthetic object datasets of overseeable size, they fail to represent larger scenes accurately, presumably due to the use of only one global latent code for encoding an entire scene or object. We propose to encode only parts of objects with features attached to unstructured point clouds. To this end we use a hierarchical feature map in 3D space, extracted from the input point clouds, with which local latent shape encodings can be queried at arbitrary positions. We use a permutohedral lattice to process the hierarchical feature maps sparsely and efficiently. This enables accurate and detailed point cloud-based reconstructions for large amounts of points in a time-efficient manner, showing good generalization capabilities across different datasets. Experiments on synthetic and real world datasets demonstrate the reconstruction capability of our method and compare favorably to state-of-the-art methods.

Recent progress in consumer hardware allowed for the collection of a large amount of animated point cloud data, which is on the one hand highly redundant and on the other hand incomplete. Our goal is to bridge this gap and find a low dimensional representation capable of approximation to a desired precision and completion of missing data. Model-less non-rigid 3D reconstruction algorithms, formulated as a linear factorization of observed point tracks into static shape component and dynamic pose, have been found insufficient to create suitable generative models, capable of generating new unobserved poses. This is due to the non-locality of the linear models, over-fitting to the non-causal correlations present in the data, which manifests in the reconstruction containing rigidly behaving not directly connected parts. In this paper, we propose a new method that can distinguish body parts and factorize the data into shape and pose purely using topological properties of the manifold-local deformations and neighborhoods. To obtain localized factorization, we formulate the deformation distance between two point tracks as the smallest deformation along the path between them. After embedding such distance in low dimensional space, a clustering of embedded data leads to close to rigid components, suitable as initialization for fitting a model-a skinned rigged mesh, used extensively in computer graphics. As both local deformations and neighborhoods of a point are local and can be estimated only from the part of the animation, the method can be used to recover unobserved data in each frame.

International audience; Visual localization enables autonomous vehicles to navigate in their surroundings and augmented reality applications to linkvirtual to real worlds. Practical visual localization approaches need to be robust to a wide variety of viewing conditions, includingday-night changes, as well as weather and seasonal variations, while providing highly accurate six degree-of-freedom (6DOF) camerapose estimates. In this paper, we extend three publicly available datasets containing images captured under a wide variety of viewingconditions, but lacking camera pose information, with ground truth pose information, making evaluation of the impact of various factorson 6DOF camera pose estimation accuracy possible. We also discuss the performance of state-of-the-art localization approaches onthese datasets. Additionally, we release around half of the poses for all conditions, and keep the remaining half private as a test set, inthe hopes that this will stimulate research on long-term visual localization, learned local image features, and related research areas.Our datasets are available at visuallocalization.net, where we are also hosting a benchmarking server for automatic evaluation ofresults on the test set. The presented state-of-the-art results are to a large degree based on submissions to our server

Modeling hand-object manipulations is essential for understanding how humans interact with their environment. While of practical importance, estimating the pose of hands and objects during interactions is challenging due to the large mutual occlusions that occur during manipulation. Recent efforts have been directed towards fully-supervised methods that require large amounts of labeled training samples. Collecting 3D ground-truth data for hand-object interactions, however, is costly, tedious, and error-prone. To overcome this challenge we present a method to leverage photometric consistency across time when annotations are only available for a sparse subset of frames in a video. Our model is trained end-to-end on color images to jointly reconstruct hands and objects in 3D by inferring their poses. Given our estimated reconstructions, we differentiably render the optical flow between pairs of adjacent images and use it within the network to warp one frame to another. We then apply a self-supervised photometric loss that relies on the visual consistency between nearby images. We achieve state-of-the-art results on 3D hand-object reconstruction benchmarks and demonstrate that our approach allows us to improve the pose estimation accuracy by leveraging information from neighboring frames in low-data regimes., Comment: CVPR 2020. See the project webpage at https://hassony2.github.io/handobjectconsist.html

We present a novel network for rolling shutter effect correction. Our network takes two consecutive rolling shutter images and estimates the corresponding global shutter image of the latest frame. The dense displacement field from a rolling shutter image to its corresponding global shutter image is estimated via a motion estimation network. The learned feature representation of a rolling shutter image is then warped, via the displacement field, to its global shutter representation by a differentiable forward warping block. An image decoder recovers the global shutter image based on the warped feature representation. Our network can be trained end-to-end and only requires the global shutter image for supervision. Since there is no public dataset available, we also propose two large datasets: the Carla-RS dataset and the Fastec-RS dataset. Experimental results demonstrate that our network outperforms the state-of-the-art methods. We make both our code and datasets available at https://github.com/ethliup/DeepUnrollNet.

Visual localization is the problem of estimating a camera within a scene and a key component in computer vision applications such as self-driving cars and Mixed Reality. State-of-the-art approaches for accurate visual localization use scene-specific representations, resulting in the overhead of constructing these models when applying the techniques to new scenes. Recently, deep learning-based approaches based on relative pose estimation have been proposed, carrying the promise of easily adapting to new scenes. However, it has been shown such approaches are currently significantly less accurate than state-of-the-art approaches. In this paper, we are interested in analyzing this behavior. To this end, we propose a novel framework for visual localization from relative poses. Using a classical feature-based approach within this framework, we show state-of-the-art performance. Replacing the classical approach with learned alternatives at various levels, we then identify the reasons for why deep learned approaches do not perform well. Based on our analysis, we make recommendations for future work., Accepted to ICRA 2020

In this paper, we present an omnidirectional localization and dense mapping system for a wide-baseline multiview stereo setup with ultra-wide field-of-view (FOV) fisheye cameras, which has a 360 degrees coverage of stereo observations of the environment. For more practical and accurate reconstruction, we first introduce improved and light-weighted deep neural networks for the omnidirectional depth estimation, which are faster and more accurate than the existing networks. Second, we integrate our omnidirectional depth estimates into the visual odometry (VO) and add a loop closing module for global consistency. Using the estimated depth map, we reproject keypoints onto each other view, which leads to a better and more efficient feature matching process. Finally, we fuse the omnidirectional depth maps and the estimated rig poses into the truncated signed distance function (TSDF) volume to acquire a 3D map. We evaluate our method on synthetic datasets with ground-truth and real-world sequences of challenging environments, and the extensive experiments show that the proposed system generates excellent reconstruction results in both synthetic and real-world environments., accepted by ICRA 2020

Many geometric quantities can be computed efficiently for convex meshes. For general meshes, methods for approximate convex decomposition have been developed that decompose a static, non-convex object into a small set of approximately convex parts. The convex hulls of those parts can then be used as a piecewise convex approximation to the original mesh. While previous work was only concerned with static meshes, we present a method for decomposing animated 3D meshes into temporally coherent approximately convex parts. Given a mesh and several training frames---that is, different spatial configurations of its vertices---we precompute an approximate convex decomposition that is independent of any specific frame. Such a decomposition can be transferred in real-time to novel, unseen frames. We apply our method to a variety of pre-animated meshes as well as a 3D character interactively controlled by a user's body pose. We further demonstrate that our method enables real-time physics simulations to interact with animated meshes.

Accurate representation of soft transitions between image regions is essential for high-quality image editing and compositing. Current techniques for generating such representations depend heavily on interaction by a skilled visual artist, as creating such accurate object selections is a tedious task. In this work, we introduce semantic soft segments , a set of layers that correspond to semantically meaningful regions in an image with accurate soft transitions between different objects. We approach this problem from a spectral segmentation angle and propose a graph structure that embeds texture and color features from the image as well as higher-level semantic information generated by a neural network. The soft segments are generated via eigendecomposition of the carefully constructed Laplacian matrix fully automatically. We demonstrate that otherwise complex image editing tasks can be done with little effort using semantic soft segments.

In this paper we review current state-of-the-art in 3D point cloud classification, present a new 3D point cloud classification benchmark data set of single scans with over four billion manually labeled points, and discuss first available results on the benchmark. Much of the stunning recent progress in 2D image interpretation can be attributed to the availability of large amounts of training data, which have enabled the (supervised) learning of deep neural networks. With the data set presented in this paper, we aim to boost the performance of CNNs also for 3D point cloud labeling. Our hope is that this will lead to a breakthrough of deep learning also for 3D (geo-) data. The semantic3D.net data set consists of dense point clouds acquired with static terrestrial laser scanners. It contains eight semantic classes and covers a wide range of urban outdoor scenes, including churches, streets, railroad tracks, squares, villages, soccer fields, and castles. We describe our labeling interface and show that, compared to those already available to the research community, our data set provides denser and more complete point clouds, with a much higher overall number of labeled points. We further provide descriptions of baseline methods and of the first independent submissions, which are indeed based on CNNs, and already show remarkable improvements over prior art. We hope that semantic3D.net will pave the way for deep learning in 3D point cloud analysis, and for 3D representation learning in general.

Comparison among non-central systems for single-view line metric reconstruction.Non-Manhattan line metric reconstruction from single image in non-central panoramas.Automatic line-image extraction in non-central panoramas. In certain non-central imaging systems, straight lines are projected via a non-planar surface encapsulating the 4 degrees of freedom of the 3D line. Consequently the geometry of the 3D line can be recovered from a minimum of four image points. However, with classical non-central catadioptric systems there is not enough effective baseline for a practical implementation of the method. In this paper we propose a multi-camera system configuration resembling the circular panoramic model which results in a particular non-central projection allowing the stitching of a non-central panorama. From a single panorama we obtain well-conditioned 3D reconstruction of lines, which are specially interesting in texture-less scenarios. No previous information about the direction or arrangement of the lines in the scene is assumed. The proposed method is evaluated on both synthetic and real images.

This paper presents a new 3D point cloud classification benchmark data set with over four billion manually labelled points, meant as input for data-hungry (deep) learning methods. We also discuss first submissions to the benchmark that use deep convolutional neural networks (CNNs) as a work horse, which already show remarkable performance improvements over state-of-the-art. CNNs have become the de-facto standard for many tasks in computer vision and machine learning like semantic segmentation or object detection in images, but have no yet led to a true breakthrough for 3D point cloud labelling tasks due to lack of training data. With the massive data set presented in this paper, we aim at closing this data gap to help unleash the full potential of deep learning methods for 3D labelling tasks. Our semantic3D.net data set consists of dense point clouds acquired with static terrestrial laser scanners. It contains 8 semantic classes and covers a wide range of urban outdoor scenes: churches, streets, railroad tracks, squares, villages, soccer fields and castles. We describe our labelling interface and show that our data set provides more dense and complete point clouds with much higher overall number of labelled points compared to those already available to the research community. We further provide baseline method descriptions and comparison between methods submitted to our online system. We hope semantic3D.net will pave the way for deep learning methods in 3D point cloud labelling to learn richer, more general 3D representations, and first submissions after only a few months indicate that this might indeed be the case., Accepted to ISPRS Annals. The benchmark website is available at http://www.semantic3d.net/ . The baseline code is available at https://github.com/nsavinov/semantic3dnet

Approach for reconstructing large-scale outdoor scenes using monocular motion stereo.Runs at interactive frame rates on a Google Project Tango Development Kit Tablet.Can use strict outlier filtering, as the user can interactively record more data.Missing parts in the live preview guide the user to places which require more data. This paper presents an approach for reconstructing large-scale outdoor scenes through monocular motion stereo at interactive frame rates on a modern mobile device (Google Project Tango Development Kit Tablet). The devices fisheye camera enables a user to reconstruct large scenes in only a few minutes by simply walking through the scene. We utilize the devices GPU to compute depth maps via plane sweep stereo. In contrast to reconstructing small objects, we observe that in large-scale scenarios using motion stereo, free-space measurements are less effective for suppressing outliers due to limited possibilities for camera placement and an unbounded reconstruction volume. Furthermore, the outlier ratio in depth maps from stereo matching is much higher compared to images from depth sensors. Consequently, we propose a set of filtering steps to detect and discard unreliable depth measurements. The remaining parts of the depth maps are then integrated into a volumetric representation of the scene using a truncated signed distance function. Ours is the first method to enable live reconstruction of large outdoor scenes on a mobile device. We extensively evaluate our approach, demonstrating the benefit of rigorously filtering depth maps.

We present a new method for decomposing an image into a set of soft color segments that are analogous to color layers with alpha channels that have been commonly utilized in modern image manipulation software. We show that the resulting decomposition serves as an effective intermediate image representation, which can be utilized for performing various, seemingly unrelated, image manipulation tasks. We identify a set of requirements that soft color segmentation methods have to fulfill, and present an in-depth theoretical analysis of prior work. We propose an energy formulation for producing compact layers of homogeneous colors and a color refinement procedure, as well as a method for automatically estimating a statistical color model from an image. This results in a novel framework for automatic and high-quality soft color segmentation that is efficient, parallelizable, and scalable. We show that our technique is superior in quality compared to previous methods through quantitative analysis as well as visually through an extensive set of examples. We demonstrate that our soft color segments can easily be exported to familiar image manipulation software packages and used to produce compelling results for numerous image manipulation applications without forcing the user to learn new tools and workflows.

We present a super-resolution method capable of creating a high-resolution texture map for a virtual 3D object from a set of lower-resolution images of that object. Our architecture unifies the concepts of (i) multi-view super-resolution based on the redundancy of overlapping views and (ii) single-view super-resolution based on a learned prior of high-resolution (HR) image structure. The principle of multi-view super-resolution is to invert the image formation process and recover the latent HR texture from multiple lower-resolution projections. We map that inverse problem into a block of suitably designed neural network layers, and combine it with a standard encoder-decoder network for learned single-image super-resolution. Wiring the image formation model into the network avoids having to learn perspective mapping from textures to images, and elegantly handles a varying number of input views. Experiments demonstrate that the combination of multi-view observations and learned prior yields improved texture maps., 11 pages, 5 figures, 2019 International Conference on 3D Vision (3DV)

Local feature matching is a critical part of many computer vision pipelines, including among others Structure-from-Motion, SLAM, and Visual Localization. However, due to limitations in the descriptors, raw matches are often contaminated by a majority of outliers. As a result, outlier detection is a fundamental problem in computer vision and a wide range of approaches, from simple checks based on descriptor similarity to geometric verification, have been proposed over the last decades. In recent years, deep learning-based approaches to outlier detection have become popular. Unfortunately, the corresponding works rarely compare with strong classical baselines. In this paper we revisit handcrafted approaches to outlier filtering. Based on best practices, we propose a hierarchical pipeline for effective outlier detection as well as integrate novel ideas which in sum lead to an efficient and competitive approach to outlier rejection. We show that our approach, although not relying on learning, is more than competitive to both recent learned works as well as handcrafted approaches, both in terms of efficiency and effectiveness. The code is available at https://github.com/cavalli1234/AdaLAM.

The efficient fusion of depth maps is a key part of most state-of-the-art 3D reconstruction methods. Besides requiring high accuracy, these depth fusion methods need to be scalable and real-time capable. To this end, we present a novel real-time capable machine learning-based method for depth map fusion. Similar to the seminal depth map fusion approach by Curless and Levoy, we only update a local group of voxels to ensure real-time capability. Instead of a simple linear fusion of depth information, we propose a neural network that predicts non-linear updates to better account for typical fusion errors. Our network is composed of a 2D depth routing network and a 3D depth fusion network which efficiently handle sensor-specific noise and outliers. This is especially useful for surface edges and thin objects for which the original approach suffers from thickening artifacts. Our method outperforms the traditional fusion approach and related learned approaches on both synthetic and real data. We demonstrate the performance of our method in reconstructing fine geometric details from noise and outlier contaminated data on various scenes., Comment: 11 pages, 8 figures, accepted to CVPR 2020

Multi-sensor fusion of multi-modal measurements from commodity inertial, visual and LiDAR sensors to provide robust and accurate 6DOF pose estimation holds great potential in robotics and beyond. In this paper, building upon our prior work (i.e., LIC-Fusion), we develop a sliding-window filter based LiDAR-Inertial-Camera odometry with online spatiotemporal calibration (i.e., LIC-Fusion 2.0), which introduces a novel sliding-window plane-feature tracking for efficiently processing 3D LiDAR point clouds. In particular, after motion compensation for LiDAR points by leveraging IMU data, low-curvature planar points are extracted and tracked across the sliding window. A novel outlier rejection criterion is proposed in the plane-feature tracking for high-quality data association. Only the tracked planar points belonging to the same plane will be used for plane initialization, which makes the plane extraction efficient and robust. Moreover, we perform the observability analysis for the LiDAR-IMU subsystem and report the degenerate cases for spatiotemporal calibration using plane features. While the estimation consistency and identified degenerate motions are validated in Monte-Carlo simulations, different real-world experiments are also conducted to show that the proposed LIC-Fusion 2.0 outperforms its predecessor and other state-of-the-art methods., Comment: 8 pages; Accepted by a conference

On the pursuit of autonomous flying robots, the scientific community has been developing onboard real-time algorithms for localisation, mapping and planning. Despite recent progress, the available solutions still lack accuracy and robustness in many aspects. While mapping for autonomous cars had a substantive boost using deep-learning techniques to enhance LIDAR measurements using image-based depth completion, the large viewpoint variations experienced by aerial vehicles are still posing major challenges for learning-based mapping approaches. In this paper, we propose a depth completion and uncertainty estimation approach that better handles the challenges of aerial platforms, such as large viewpoint and depth variations, and limited computing resources. The core of our method is a novel compact network that performs both depth completion and confidence estimation using an image-guided approach. Real-time performance onboard a GPU suitable for small flying robots is achieved by sharing deep features between both tasks. Experiments demonstrate that our network outperforms the state-of-the-art in depth completion and uncertainty estimation for single-view methods on mobile GPUs. We further present a new photorealistic aerial depth completion dataset that exhibits more challenging depth completion scenarios than the established indoor and car driving datasets. The dataset includes an open-source, visual-inertial UAV simulator for photo-realistic data generation. Our results show that our network trained on this dataset can be directly deployed on real-world outdoor aerial public datasets without fine-tuning or style transfer., IEEE Robotics and Automation Letters, 5 (2), ISSN:2377-3766

In this work, we address the problem of refining the geometry of local image features from multiple views without known scene or camera geometry. Current approaches to local feature detection are inherently limited in their keypoint localization accuracy because they only operate on a single view. This limitation has a negative impact on downstream tasks such as Structure-from-Motion, where inaccurate keypoints lead to large errors in triangulation and camera localization. Our proposed method naturally complements the traditional feature extraction and matching paradigm. We first estimate local geometric transformations between tentative matches and then optimize the keypoint locations over multiple views jointly according to a non-linear least squares formulation. Throughout a variety of experiments, we show that our method consistently improves the triangulation and camera localization performance for both hand-crafted and learned local features.

To be invariant, or not to be invariant: that is the question formulated in this work about local descriptors. A limitation of current feature descriptors is the trade-off between generalization and discriminative power: more invariance means less informative descriptors. We propose to overcome this limitation with a disentanglement of invariance in local descriptors and with an online selection of the most appropriate invariance given the context. Our framework (https://github.com/rpautrat/LISRD) consists in a joint learning of multiple local descriptors with different levels of invariance and of meta descriptors encoding the regional variations of an image. The similarity of these meta descriptors across images is used to select the right invariance when matching the local descriptors. Our approach, named Local Invariance Selection at Runtime for Descriptors (LISRD), enables descriptors to adapt to adverse changes in images, while remaining discriminative when invariance is not required. We demonstrate that our method can boost the performance of current descriptors and outperforms state-of-the-art descriptors in several matching tasks, when evaluated on challenging datasets with day-night illumination as well as viewpoint changes.

We address the problem of mesh reconstruction from live RGB-D video, assuming a calibrated camera and poses provided externally (e.g., by a SLAM system). In contrast to most existing approaches, we do not fuse depth measurements in a volume but in a dense surfel cloud. We asynchronously (re)triangulate the smoothed surfels to reconstruct a surface mesh. This novel approach enables to maintain a dense surface representation of the scene during SLAM which can quickly adapt to loop closures. This is possible by deforming the surfel cloud and asynchronously remeshing the surface where necessary. The surfel-based representation also naturally supports strongly varying scan resolution. In particular, it reconstructs colors at the input camera's resolution. Moreover, in contrast to many volumetric approaches, ours can reconstruct thin objects since objects do not need to enclose a volume. We demonstrate our approach in a number of experiments, showing that it produces reconstructions that are competitive with the state-of-the-art, and we discuss its advantages and limitations. The algorithm (excluding loop closure functionality) is available as open source at https://github.com/puzzlepaint/surfelmeshing ., Comment: Version accepted to IEEE Transactions on Pattern Analysis and Machine Intelligence

To model radial distortion there are two main approaches; either the image points are undistorted such that they correspond to pinhole projections, or the pinhole projections are distorted such that they align with the image measurements. Depending on the application, either of the two approaches can be more suitable. For example, distortion models are commonly used in Structure-from-Motion since they simplify measuring the reprojection error in images. Surprisingly, all previous minimal solvers for pose estimation with radial distortion use undistortion models. In this paper we aim to fill this gap in the literature by proposing the first minimal solvers which can jointly estimate distortion models together with camera pose. We present a general approach which can handle rational models of arbitrary degree for both distortion and undistortion.

Augmented/mixed reality and robotic applications are increasingly relying on cloud-based localization services, which require users to upload query images to perform camera pose estimation on a server. This raises significant privacy concerns when consumers use such services in their homes or in confidential industrial settings. Even if only image features are uploaded, the privacy concerns remain as the images can be reconstructed fairly well from feature locations and descriptors. We propose to conceal the content of the query images from an adversary on the server or a man-in-the-middle intruder. The key insight is to replace the 2D image feature points in the query image with randomly oriented 2D lines passing through their original 2D positions. It will be shown that this feature representation hides the image contents, and thereby protects user privacy, yet still provides sufficient geometric constraints to enable robust and accurate 6-DOF camera pose estimation from feature correspondences. Our proposed method can handle single- and multi-image queries as well as exploit additional information about known structure, gravity, and scale. Numerous experiments demonstrate the high practical relevance of our approach.

In this paper we consider the problem of relative pose estimation from two images with per-pixel polarimetric information. Using these additional measurements we derive a simple minimal solver for the essential matrix which only requires two point correspondences. The polarization constraints allow us to pointwise recover the 3D surface normal up to a two-fold ambiguity for the diffuse reflection. Since this ambiguity exists per point, there is a combinatorial explosion of possibilities. However, since our solver only requires two point correspondences, we only need to consider 16 configurations when solving for the relative pose. Once the relative orientation is recovered, we show that it is trivial to resolve the ambiguity for the remaining points. For robustness, we also propose a joint optimization between the relative pose and the refractive index to handle the refractive distortion. In experiments, on both synthetic and real data, we demonstrate that by leveraging the additional information available from polarization cameras, we can improve over classical methods which only rely on the 2D-point locations to estimate the geometry. Finally, we demonstrate the practical applicability of our approach by integrating it into a state-of-the-art global Structure-from-Motion pipeline.

We propose a novel method for instance label segmentation of dense 3D voxel grids. We target volumetric scene representations, which have been acquired with depth sensors or multi-view stereo methods and which have been processed with semantic 3D reconstruction or scene completion methods. The main task is to learn shape information about individual object instances in order to accurately separate them, including connected and incompletely scanned objects. We solve the 3D instance-labeling problem with a multi-task learning strategy. The first goal is to learn an abstract feature embedding, which groups voxels with the same instance label close to each other while separating clusters with different instance labels from each other. The second goal is to learn instance information by densely estimating directional information of the instance's center of mass for each voxel. This is particularly useful to find instance boundaries in the clustering post-processing step, as well as, for scoring the segmentation quality for the first goal. Both synthetic and real-world experiments demonstrate the viability and merits of our approach. In fact, it achieves state-of-the-art performance on the ScanNet 3D instance segmentation benchmark.

In this paper, we present a method to utilize 2D-2D point matches between images taken during different image conditions to train a convolutional neural network for semantic segmentation. Enforcing label consistency across the matches makes the final segmentation algorithm robust to seasonal changes. We describe how these 2D-2D matches can be generated with little human interaction by geometrically matching points from 3D models built from images. Two cross-season correspondence datasets are created providing 2D-2D matches across seasonal changes as well as from day to night. The datasets are made publicly available to facilitate further research. We show that adding the correspondences as extra supervision during training improves the segmentation performance of the convolutional neural network, making it more robust to seasonal changes and weather conditions., Comment: In Proc. CVPR 2019

A key component of Simultaneous Localization and Mapping (SLAM) systems is the joint optimization of the estimated 3D map and camera trajectory. Bundle adjustment (BA) is the gold standard for this. Due to the large number of variables in dense RGB-D SLAM, previous work has focused on approximating BA. In contrast, in this paper we present a novel, fast direct BA formulation which we implement in a real-time dense RGB-D SLAM algorithm. In addition, we show that direct RGB-D SLAM systems are highly sensitive to rolling shutter, RGB and depth sensor synchronization, and calibration errors. In order to facilitate state-of-the-art research on direct RGB-D SLAM, we propose a novel, well-calibrated benchmark for this task that uses synchronized global shutter RGB and depth cameras. It includes a training set, a test set without public ground truth, and an online evaluation service. We observe that the ranking of methods changes on this dataset compared to existing ones, and our proposed algorithm outperforms all other evaluated SLAM methods. Our benchmark and our open source SLAM algorithm are available at: www.eth3d.net

Visual localization is the task of accurate camera pose estimation in a known scene. It is a key problem in computer vision and robotics, with applications including self-driving cars, Structure-from-Motion, SLAM, and Mixed Reality. Traditionally, the localization problem has been tackled using 3D geometry. Recently, end-to-end approaches based on convolutional neural networks have become popular. These methods learn to directly regress the camera pose from an input image. However, they do not achieve the same level of pose accuracy as 3D structure-based methods. To understand this behavior, we develop a theoretical model for camera pose regression. We use our model to predict failure cases for pose regression techniques and verify our predictions through experiments. We furthermore use our model to show that pose regression is more closely related to pose approximation via image retrieval than to accurate pose estimation via 3D structure. A key result is that current approaches do not consistently outperform a handcrafted image retrieval baseline. This clearly shows that additional research is needed before pose regression algorithms are ready to compete with structure-based methods., Initial version of a paper accepted to CVPR 2019

Localizing an image wrt. a 3D scene model represents a core task for many computer vision applications. An increasing number of real-world applications of visual localization on mobile devices, e.g., Augmented Reality or autonomous robots such as drones or self-driving cars, demand localization approaches to minimize storage and bandwidth requirements. Compressing the 3D models used for localization thus becomes a practical necessity. In this work, we introduce a new hybrid compression algorithm that uses a given memory limit in a more effective way. Rather than treating all 3D points equally, it represents a small set of points with full appearance information and an additional, larger set of points with compressed information. This enables our approach to obtain a more complete scene representation without increasing the memory requirements, leading to a superior performance compared to previous compression schemes. As part of our contribution, we show how to handle ambiguous matches arising from point compression during RANSAC. Besides outperforming previous compression techniques in terms of pose accuracy under the same memory constraints, our compression scheme itself is also more efficient. Furthermore, the localization rates and accuracy obtained with our approach are comparable to state-of-the-art feature-based methods, while using a small fraction of the memory., Comment: Published at CVPR 2019

We tackle the problem of retrieving high-resolution (HR) texture maps of objects that are captured from multiple view points. In the multi-view case, model-based super-resolution (SR) methods have been recently proved to recover high quality texture maps. On the other hand, the advent of deep learning-based methods has already a significant impact on the problem of video and image SR. Yet, a deep learning-based approach to super-resolve the appearance of 3D objects is still missing. The main limitation of exploiting the power of deep learning techniques in the multi-view case is the lack of data. We introduce a 3D appearance SR (3DASR) dataset based on the existing ETH3D [42], SyB3R [31], MiddleBury, and our Collection of 3D scenes from TUM [21], Fountain [51] and Relief [53]. We provide the high- and low-resolution texture maps, the 3D geometric model, images and projection matrices. We exploit the power of 2D learning-based SR methods and design networks suitable for the 3D multi-view case. We incorporate the geometric information by introducing normal maps and further improve the learning process. Experimental results demonstrate that our proposed networks successfully incorporate the 3D geometric information and super-resolve the texture maps., In CVPR 2019. Github papge: https://github.com/ofsoundof/3D_Appearance_SR

We present a stereo-based dense mapping algorithm for large-scale dynamic urban environments. In contrast to other existing methods, we simultaneously reconstruct the static background, the moving objects, and the potentially moving but currently stationary objects separately, which is desirable for high-level mobile robotic tasks such as path planning in crowded environments. We use both instance-aware semantic segmentation and sparse scene flow to classify objects as either background, moving, or potentially moving, thereby ensuring that the system is able to model objects with the potential to transition from static to dynamic, such as parked cars. Given camera poses estimated from visual odometry, both the background and the (potentially) moving objects are reconstructed separately by fusing the depth maps computed from the stereo input. In addition to visual odometry, sparse scene flow is also used to estimate the 3D motions of the detected moving objects, in order to reconstruct them accurately. A map pruning technique is further developed to improve reconstruction accuracy and reduce memory consumption, leading to increased scalability. We evaluate our system thoroughly on the well-known KITTI dataset. Our system is capable of running on a PC at approximately 2.5Hz, with the primary bottleneck being the instance-aware semantic segmentation, which is a limitation we hope to address in future work. The source code is available from the project website (http://andreibarsan.github.io/dynslam)., Presented at IEEE International Conference on Robotics and Automation (ICRA), 2018

Visual localization is a key step in many robotics pipelines, allowing the robot to (approximately) determine its position and orientation in the world. An efficient and scalable approach to visual localization is to use image retrieval techniques. These approaches identify the image most similar to a query photo in a database of geo-tagged images and approximate the query's pose via the pose of the retrieved database image. However, image retrieval across drastically different illumination conditions, e.g. day and night, is still a problem with unsatisfactory results, even in this age of powerful neural models. This is due to a lack of a suitably diverse dataset with true correspondences to perform end-to-end learning. A recent class of neural models allows for realistic translation of images among visual domains with relatively little training data and, most importantly, without ground-truth pairings. In this paper, we explore the task of accurately localizing images captured from two traversals of the same area in both day and night. We propose ToDayGAN - a modified image-translation model to alter nighttime driving images to a more useful daytime representation. We then compare the daytime and translated night images to obtain a pose estimate for the night image using the known 6-DOF position of the closest day image. Our approach improves localization performance by over 250% compared the current state-of-the-art, in the context of standard metrics in multiple categories., Comment: Published in ICRA 2019

We present a unified framework for understanding 3D hand and object interactions in raw image sequences from egocentric RGB cameras. Given a single RGB image, our model jointly estimates the 3D hand and object poses, models their interactions, and recognizes the object and action classes with a single feed-forward pass through a neural network. We propose a single architecture that does not rely on external detection algorithms but rather is trained end-to-end on single images. We further merge and propagate information in the temporal domain to infer interactions between hand and object trajectories and recognize actions. The complete model takes as input a sequence of frames and outputs per-frame 3D hand and object pose predictions along with the estimates of object and action categories for the entire sequence. We demonstrate state-of-the-art performance of our algorithm even in comparison to the approaches that work on depth data and ground-truth annotations., CVPR 2019 (Oral)

This paper discusses a reconstruction challenge held as a part of the second 3D Reconstruction meets Semantics workshop (3DRMS). The challenge goals and datasets are introduced, including both synthetic and real data from outdoor scenes, here represented by gardens with a variety of bushes, trees, other plants and objects. Both qualitative and quantitative evaluation of the challenge participants’ submissions is given in categories of geometric and semantic accuracy. Finally, comparison of submitted results with baseline methods is given, showing a modest performance increase in some of the categories.

We present a fully convolutional neural network that jointly predicts a semantic 3D reconstruction of a scene as well as a corresponding octree representation. This approach leverages the efficiency of an octree data structure to improve the capacities of volumetric semantic 3D reconstruction methods, especially in terms of scalability. At every octree level, the network predicts a semantic class for every voxel and decides which voxels should be further split in order to refine the reconstruction, thus working in a coarse-to-fine manner. The semantic prediction part of our method builds on recent work that combines traditional variational optimization and neural networks. In contrast to previous networks that work on dense voxel grids, our network is much more efficient in terms of memory consumption and inference efficiency, while achieving similar reconstruction performance. This allows for a high resolution reconstruction in case of limited memory. We perform experiments on the SUNCG and ScanNetv2 datasets on which our network shows comparable reconstruction results to the corresponding dense network while consuming less memory.

Camera calibration is an essential first step in setting up 3D Computer Vision systems. Commonly used parametric camera models are limited to a few degrees of freedom and thus often do not optimally fit to complex real lens distortion. In contrast, generic camera models allow for very accurate calibration due to their flexibility. Despite this, they have seen little use in practice. In this paper, we argue that this should change. We propose a calibration pipeline for generic models that is fully automated, easy to use, and can act as a drop-in replacement for parametric calibration, with a focus on accuracy. We compare our results to parametric calibrations. Considering stereo depth estimation and camera pose estimation as examples, we show that the calibration error acts as a bias on the results. We thus argue that in contrast to current common practice, generic models should be preferred over parametric ones whenever possible. To facilitate this, we released our calibration pipeline at https://github.com/puzzlepaint/camera_calibration, making both easy-to-use and accurate camera calibration available to everyone., Comment: 15 pages, 12 figures, accepted to CVPR 2020 as an oral

This work presents and evaluates a novel compact scene representation based on Stixels that infers geometric and semantic information. Our approach overcomes the previous rather restrictive geometric assumptions for Stixels by introducing a novel depth model to account for non-flat roads and slanted objects. Both semantic and depth cues are used jointly to infer the scene representation in a sound global energy minimization formulation. Furthermore, a novel approximation scheme is introduced in order to significantly reduce the computational complexity of the Stixel algorithm, and then achieve real-time computation capabilities. The idea is to first perform an over-segmentation of the image, discarding the unlikely Stixel cuts, and apply the algorithm only on the remaining Stixel cuts. This work presents a novel over-segmentation strategy based on a fully convolutional network, which outperforms an approach based on using local extrema of the disparity map. We evaluate the proposed methods in terms of semantic and geometric accuracy as well as run-time on four publicly available benchmark datasets. Our approach maintains accuracy on flat road scene datasets while improving substantially on a novel non-flat road dataset., International Journal of Computer Vision, 127 (11-12), ISSN:0920-5691, ISSN:1573-1405

We present a visual odometry (VO) algorithm for a multi-camera system and robust operation in challenging environments. Our algorithm consists of a pose tracker and a local mapper. The tracker estimates the current pose by minimizing photometric errors between the most recent keyframe and the current frame. The mapper initializes the depths of all sampled feature points using plane-sweeping stereo. To reduce pose drift, a sliding window optimizer is used to refine poses and structure jointly. Our formulation is flexible enough to support an arbitrary number of stereo cameras. We evaluate our algorithm thoroughly on five datasets. The datasets were captured in different conditions: daytime, night-time with near-infrared (NIR) illumination and nighttime without NIR illumination. Experimental results show that a multi-camera setup makes the VO more robust to challenging environments, especially night-time conditions, in which a single stereo configuration fails easily due to the lack of features.

Project AutoVision aims to develop localization and 3D scene perception capabilities for a self-driving vehicle. Such capabilities will enable autonomous navigation in urban and rural environments, in day and night, and with cameras as the only exteroceptive sensors. The sensor suite employs many cameras for both 360-degree coverage and accurate multi-view stereo; the use of low-cost cameras keeps the cost of this sensor suite to a minimum. In addition, the project seeks to extend the operating envelope to include GNSS-less conditions which are typical for environments with tall buildings, foliage, and tunnels. Emphasis is placed on leveraging multi-view geometry and deep learning to enable the vehicle to localize and perceive in 3D space. This paper presents an overview of the project, and describes the sensor suite and current progress in the areas of calibration, localization, and perception.

In this paper, we aim to solve the pose estimation problem of calibrated pinhole and generalized cameras w.r.t. a Structure-from-Motion (SfM) model by leveraging both 2D-3D correspondences as well as 2D-2D correspondences. Traditional approaches either focus on the use of 2D-3D matches, known as structure-based pose estimation or solely on 2D-2D matches (structure-less pose estimation). Absolute pose approaches are limited in their performance by the quality of the 3D point triangulations as well as the completeness of the 3D model. Relative pose approaches, on the other hand, while being more accurate, also tend to be far more computationally costly and often return dozens of possible solutions. This work aims to bridge the gap between these two paradigms. We propose a new RANSAC-based approach that automatically chooses the best type of solver to use at each iteration in a data-driven way. The solvers chosen by our RANSAC can range from pure structure-based or structure-less solvers, to any possible combination of hybrid solvers (i.e. using both types of matches) in between. A number of these new hybrid minimal solvers are also presented in this paper. Both synthetic and real data experiments show our approach to be as accurate as structure-less approaches, while staying close to the efficiency of structure-based methods.

Visual localization enables autonomous vehicles to navigate in their surroundings and augmented reality applications to link virtual to real worlds. Practical visual localization approaches need to be robust to a wide variety of viewing condition, including day-night changes, as well as weather and seasonal variations, while providing highly accurate 6 degree-of-freedom (6DOF) camera pose estimates. In this paper, we introduce the first benchmark datasets specifically designed for analyzing the impact of such factors on visual localization. Using carefully created ground truth poses for query images taken under a wide variety of conditions, we evaluate the impact of various factors on 6DOF camera pose estimation accuracy through extensive experiments with state-of-the-art localization approaches. Based on our results, we draw conclusions about the difficulty of different conditions, showing that long-term localization is far from solved, and propose promising avenues for future work, including sequence-based localization approaches and the need for better local features. Our benchmark is available at visuallocalization.net.

We propose a novel method for the geometric registration of semantically labeled regions. We approximate semantic regions by ellipsoids, and leverage their convexity to formulate the correspondence search effectively as a constrained optimization problem that maximizes the number of matched regions, and which we solve globally optimal in a Branch-and-Bound fashion. To this end, we derive suitable linear matrix inequality constraints which describe ellipsoid-to-ellipsoid assignment conditions. Our approach is robust to large percentages of outliers and thus applicable to difficult correspondence search problems. In multiple experiments we demonstrate the flexibility and robustness of our approach on a number of challenging vision problems.

Illumination is a critical element of photography and is essential for many computer vision tasks. Flash light is unique in the sense that it is a widely available tool for easily manipulating the scene illumination. We present a dataset of thousands of ambient and flash illumination pairs to enable studying flash photography and other applications that can benefit from having separate illuminations. Different than the typical use of crowdsourcing in generating computer vision datasets, we make use of the crowd to directly take the photographs that make up our dataset. As a result, our dataset covers a wide variety of scenes captured by many casual photographers. We detail the advantages and challenges of our approach to crowdsourcing as well as the computational effort to generate completely separate flash illuminations from the ambient light in an uncontrolled setup. We present a brief examination of illumination decomposition, a challenging and underconstrained problem in flash photography, to demonstrate the use of our dataset in a data-driven approach.

Visual localization, i.e., determining the position and orientation of a vehicle with respect to a map, is a key problem in autonomous driving. We present a multicamera visual inertial localization algorithm for large scale environments. To efficiently and effectively match features against a pre-built global 3D map, we propose a prioritized feature matching scheme for multi-camera systems. In contrast to existing works, designed for monocular cameras, we (1) tailor the prioritization function to the multi-camera setup and (2) run feature matching and pose estimation in parallel. This significantly accelerates the matching and pose estimation stages and allows us to dynamically adapt the matching efforts based on the surrounding environment. In addition, we show how pose priors can be integrated into the localization system to increase efficiency and robustness. Finally, we extend our algorithm by fusing the absolute pose estimates with motion estimates from a multi-camera visual inertial odometry pipeline (VIO). This results in a system that provides reliable and drift-less pose estimation. Extensive experiments show that our localization runs fast and robust under varying conditions, and that our extended algorithm enables reliable real-time pose estimation., Comment: 7 pages, 5 figures

Robust and accurate visual localization across large appearance variations due to changes in time of day, seasons, or changes of the environment is a challenging problem which is of importance to application areas such as navigation of autonomous robots. Traditional feature-based methods often struggle in these conditions due to the significant number of erroneous matches between the image and the 3D model. In this paper, we present a method for scoring the individual correspondences by exploiting semantic information about the query image and the scene. In this way, erroneous correspondences tend to get a low semantic consistency score, whereas correct correspondences tend to get a high score. By incorporating this information in a standard localization pipeline, we show that the localization performance can be significantly improved compared to the state-of-the-art, as evaluated on two challenging long-term localization benchmarks.

We present a novel semantic 3D reconstruction framework which embeds variational regularization into a neural network. Our network performs a fixed number of unrolled multi-scale optimization iterations with shared interaction weights. In contrast to existing variational methods for semantic 3D reconstruction, our model is end-to-end trainable and captures more complex dependencies between the semantic labels and the 3D geometry. Compared to previous learning-based approaches to 3D reconstruction, we integrate powerful long-range dependencies using variational coarse-to-fine optimization. As a result, our network architecture requires only a moderate number of parameters while keeping a high level of expressiveness which enables learning from very little data. Experiments on real and synthetic datasets demonstrate that our network achieves higher accuracy compared to a purely variational approach while at the same time requiring two orders of magnitude less iterations to converge. Moreover, our approach handles ten times more semantic class labels using the same computational resources.