Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Leif P. KobbeltSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works162 from public data
- 1,281
- D2-Net: A Trainable CNN for Joint Description and Detection of Local Features1,188
This work proposes an approach where a single convolutional neural network plays a dual role: It is simultaneously a dense feature descriptor and a feature detector, and shows that this model can be trained using pixel correspondences extracted from readily available large-scale SfM reconstructions, without any further annotations.
- Mip-Splatting: Alias-Free 3D Gaussian Splatting968
A 3D smoothing filter is introduced to constrains the size of the 3D Gaussian primitives based on the maximal sampling frequency induced by the input views, which eliminates high-frequency artifacts when zooming in.
- D2-Net: A Trainable CNN for Joint Detection and Description of Local Features954
This work proposes an approach where a single convolutional neural network plays a dual role: It is simultaneously a dense feature descriptor and a feature detector, and shows that this model can be trained using pixel correspondences extracted from readily available large-scale SfM reconstructions, without any further annotations.
- 794
- 671
- 646
- 570
- 556
- Image-Based Localization Using LSTMs for Structured Feature Correlation555
Experimental results show the proposed CNN+LSTM architecture for camera pose regression for indoor and outdoor scenes outperforms existing deep architectures, and can localize images in hard conditions, where classic SIFT-based methods fail.
- 473
- Understanding the Limitations of CNN-Based Absolute Camera Pose Regression447
A theoretical model for camera pose regression is developed that is more closely related to pose approximation via image retrieval than to accurate pose estimation via 3D structure, and shows that additional research is needed before pose regression algorithms are ready to compete with structure-based methods.
- Gaussian Opacity Fields: Efficient Adaptive Surface Reconstruction in Unbounded Scenes362
This work presents Gaussian Opacity Fields (GOF), a novel approach for efficient, high-quality, and adaptive surface reconstruction in unbounded scenes, derived from ray-tracing-based volume rendering of 3D Gaussians, enabling direct geometry extraction from 3D Gaussians by identifying its levelset without resorting to Poisson reconstruction or TSDF fusion.
- Back to the Feature: Learning Robust Camera Localization from Pixels to Pose360
PixLoc is introduced, a scene-agnostic neural network that estimates an accurate 6-DoF pose from an image and a 3D model, based on the direct alignment of multiscale deep features, casting camera localization as metric learning.
- Image Retrieval for Image-Based Localization Revisited321
It is shown that retrieval methods using a selective voting scheme are able to outperform state-of-the-art direct matching methods and how both selective voting and correspondence computation can be accelerated by using a Hamming embedding of feature descriptors.
- 320
- Improving Image-Based Localization by Active Correspondence Search309
This work proposes a powerful pipeline for determining the pose of a query image relative to a point cloud reconstruction of a large scene consisting of more than one million 3D points, with run-times comparable or superior to the fastest state-of-the-art methods.
- Semantic Visual Localization285
This paper proposes a novel approach based on a joint 3D geometric and semantic understanding of the world, enabling it to succeed under conditions where previous approaches failed, and employs a novel generative model for descriptor learning.
- Get Out of My Lab: Large-scale, Real-Time Visual-Inertial Localization265
This paper demonstrates that large-scale, real-time pose estimation and tracking can be performed on mobile platforms with limited resources without the use of an external server by employing map and descriptor compression schemes as well as efficient search algorithms from computer vision.
- Patch2Pix: Epipolar-Guided Pixel-Level Correspondences258
This work presents Patch2Pix, a novel refinement network that refines match proposals by regressing pixel-level matches from the local regions defined by those proposals and jointly rejecting outlier matches with confidence scores.
- Night-to-Day Image Translation for Retrieval-based Localization248
This paper proposes ToDayGAN – a modified image-translation model to alter nighttime driving images to a more useful daytime representation, and improves localization performance by over 250% compared the current state-of-the-art, in the context of standard metrics in multiple categories.
- 246
- Scalable 6-DOF Localization on Mobile Devices233
This paper proposes two strategies to combine the information provided by local tracking and global localization with the speed and precision of a local pose tracker on a mobile device that are both scalable and drift-free by design and eliminates the need for loop closure.
- 3D visual perception for self-driving cars using a multi-camera system: Calibration, mapping, localization, and obstacle detection230
This pipeline is able to precisely calibrate multi-camera systems, build sparse 3D maps for visual navigation, visually localize the car with respect to these maps, generate accurate dense maps, as well as detect obstacles based on real-time depth map extraction.
- 213
- Quad-Networks: Unsupervised Learning to Rank for Interest Point Detection196
This paper is the first to propose such a formulation: training a neural network to rank points in a transformation-invariant manner, and shows that this unsupervised method performs better or on-par with baselines on two tasks.
- 196
- 189
- Human POSEitioning System (HPS): 3D Human Pose Estimation and Self-localization in Large Scenes from Body-Mounted Sensors185
The HPS method to recover the full 3D pose of a human registered with a 3D scan of the surrounding environment using wearable sensors fuses camera based self-localization with IMU-based human body tracking and shows that the optimization-based integration exploits the benefits of the two, resulting in pose accuracy free of drift.
- 173
- 167
- Reference Pose Generation for Long-term Visual Localization via Learned Features and View Synthesis165
This work proposes a semi-automated approach to generate reference poses based on feature matching between renderings of a 3D model and real images via learned features, and significantly improves the nighttime reference poses of the popular Aachen Day–Night dataset.
- A Vote-and-Verify Strategy for Fast Spatial Verification in Image Retrieval154
This paper develops a novel spatial verification approach that uses an efficient voting scheme to identify promising transformation hypotheses that are subsequently verified and refined and is able to achieve a verification accuracy similar to state-of-the-art hypothesize-and-verify approaches while providing faster runtimes than state of theart voting-based methods.
- VSO: Visual Semantic Odometry151
A novel visual semantic odometry (VSO) framework to enable medium-term continuous tracking of points using semantics and demonstrates a significant improvement over state-of-the-art baselines in the context of autonomous driving simply by integrating the authors' semantic constraints.
- Semantic Match Consistency for Long-Term Visual Localization151
This paper presents a method for scoring the individual correspondences by exploiting semantic information about the query image and the scene, and shows that the localization performance can be significantly improved compared to the state-of-the-art, as evaluated on two challenging long-term localization benchmarks.
- 150
- DGC-Net: Dense Geometric Correspondence Network146
This paper proposes a coarse-to-fine CNN-based framework that can leverage the advantages of optical flow approaches and extend them to the case of large transformations providing dense and subpixel accurate estimates and proves that the model outperforms existing dense approaches.
- Project AutoVision: Localization and 3D Scene Perception for an Autonomous Vehicle with a Multi-Camera System132
An overview of the Project AutoVision is presented, and the sensor suite and current progress in the areas of calibration, localization, and perception are described.
- Deep Visual Geo-localization Benchmark131
A new open-source benchmarking framework for Visual Geo-localization that allows to build, train, and test a wide range of commonly used ar-chitectures, with the flexibility to change individual components of a geo- localization pipeline is proposed.
- To Learn or Not to Learn: Visual Localization from Essential Matrices126
A novel framework for visual localization from relative poses is proposed, replacing the classical approach with learned alternatives at various levels, and the reasons for why deep learned approaches do not perform well are identified.
- 112
- Large-scale, real-time visual–inertial localization revisited106
This work aims to deploy localization at a global scale where one thus relies on methods using local features and sparse 3D models using a combination of priors, nearest-neighbor search, geometric match culling, and a cascaded pose candidate refinement step.
- SurfelMeshing: Online Surfel-Based Mesh Reconstruction97
This novel approach enables to maintain a dense surface representation of the scene during SLAM which can quickly adapt to loop closures and asynchronously remeshing the surface where necessary, and can reconstruct thin objects since objects do not need to enclose a volume.
- 94
- On the Limits of Pseudo Ground Truth in Visual Camera Re-localisation89
This paper analyzes two widely used re-localisation datasets and shows that evaluation outcomes indeed vary with the choice of the reference algorithm, and question common beliefs in the re- localisation literature, namely that learning-based scene coordinate regression outperforms classical feature-based methods, and that RGB-D- based methods outperform RGB-based Methods.
- Tetra-NeRF: Representing Neural Radiance Fields Using Tetrahedra80
This paper proposes to use an adaptive representation based on tetrahedra obtained by Delaunay triangulation instead of uniform subdivision or point-based representations, and shows that such a representation enables efficient training and leads to state-of-the-art results.
- 80
- A Cross-Season Correspondence Dataset for Robust Semantic Segmentation80
It is shown 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.
- Fine-Grained Segmentation Networks: Self-Supervised Segmentation for Improved Long-Term Visual Localization78
This paper proposes a novel neural network, the Fine-Grained Segmentation Network (FGSN), that can be used to provide image segmentations with a larger number of labels and can be trained in a self-supervised fashion.
- Why Having 10,000 Parameters in Your Camera Model Is Better Than Twelve76
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 is proposed, making both easy-to-use and accurate camera calibration available to everyone.
- Hybrid Scene Compression for Visual Localization76
A new hybrid compression algorithm is introduced that uses a given memory limit in a more effective way and is comparable to state-of-the-art feature-based methods, while using a small fraction of the memory.
- Large-scale outdoor 3D reconstruction on a mobile device76
This paper presents an approach for reconstructing large-scale outdoor scenes through monocular motion stereo at interactive frame rates on a modern mobile device, and is the first method to enable live reconstruction of large outdoor scenes on a mobile device.
- Benchmarking Image Retrieval for Visual Localization73
It is shown that retrieval performance on classical landmark retrieval/recognition tasks correlates only for some but not all tasks to localization performance, indicating a need for retrieval approaches specifically designed for localization tasks.
- 73
- 70
- 63
- Evaluating Local Features for Day-Night Matching62
It is found that feature detectors are severely affected by illumination changes between day and night and that there is great potential to improve both feature detectors and descriptors.
- Investigating the Role of Image Retrieval for Visual Localization61
There is a need for retrieval approaches specifically designed for localization paradigms, as this paper introduces a novel benchmark setup and investigates several definitions of “ground truth” for image retrieval and shows that retrieval performance on classical landmark retrieval or place recognition tasks correlates only for some but not all paradigm to localization performance.
- Image-based Localization with Spatial LSTMs.61
Experimental results show the proposed CNN+LSTM architecture for camera pose regression for indoor and outdoor scenes outperforms existing deep architectures, and can localize images in hard conditions, e.g., in the presence of mostly textureless surfaces.
- Incremental Visual-Inertial 3D Mesh Generation with Structural Regularities57
This work proposes to tightly couple mesh regularization and state estimation by detecting and enforcing structural regularities in a novel factor-graph formulation, and incrementally build the mesh by restricting its extent to the time-horizon of the VIO optimization.
- 57
- Beyond Controlled Environments: 3D Camera Re-localization in Changing Indoor Scenes56
This paper adapts 3RScan - a recently introduced indoor RGB-D dataset designed for object instance re-localization - to create RIO10, a new long-term camera re- localization benchmark focused on indoor scenes and explores how state-of-the-art cameraRe-localizers perform according to these metrics.
- 55
- Is This the Right Place? Geometric-Semantic Pose Verification for Indoor Visual Localization55
This paper develops multiple hand-crafted as well as a trainable approach to join into the geometric-semantic verification and shows significant improvements over state-of-the-art on a very challenging indoor dataset.
- Minimal Solvers for Generalized Pose and Scale Estimation from Two Rays and One Point51
This paper presents a problem formulation based on one point-point and two point-ray correspondences that leads to closed-form solutions that are orders of magnitude faster to compute than the current state-of-the-art, while resulting in a similar or better pose accuracy.
- 48
- Real-Time Dense Mapping for Self-Driving Vehicles using Fisheye Cameras47
The implementation is based on fisheye cameras whose large field of view benefits various computer vision applications for self-driving vehicles such as visual-inertial odometry, visual localization, and object detection, and adopts the fast object detection framework, YOLOv3, to remove potentially dynamic objects.
- 47
- Indoor-Outdoor 3D Reconstruction Alignment46
This work proposes a solution to align disconnected indoor and outdoor models of the same building into a single 3D model by leveraging semantic information, specifically window detections, in multiple scenes to obtain candidate matches from which an alignment hypothesis can be computed.
- Incremental Object Database: Building 3D Models from Multiple Partial Observations43
A system that incrementally builds a database of objects as a mobile agent traverses a scene, which allows to create and improve object models on the fly and to use these merged models to reconstruct also unobserved parts of the scene.
- 41
- 41
- 40
- Efficient 2D-3D Matching for Multi-Camera Visual Localization40
To efficiently and effectively match features against a pre-built global 3D map, this work proposes a prioritized feature matching scheme for multi-camera systems and shows how pose priors can be integrated into the localization system to increase efficiency and robustness.
- 39
- On Sampling Focal Length Values to Solve the Absolute Pose Problem39
This paper challenges the notion that using minimal solvers is always optimal and proposes to compute the pose for a camera with unknown focal length by randomly sampling a focal length value and using an efficient pose solver for the now calibrated camera.
- 38
- 36
- 35
- Semantically Informed Multiview Surface Refinement33
A method to jointly refine the geometry and semantic segmentation of 3D surface meshes and introduces priors that favor adaptive smoothing, depending on the class label; straightness of class boundaries; and semantic labels that are consistent with the surface orientation.
- Interaction Replica: Tracking Human–Object Interaction and Scene Changes From Human Motion32
The method, iReplica (Interaction Replica), is an essential first step towards the egocentric capture of human interactions and modeling of dynamic scenes, which is required for future AR/VR applications in immersive virtual universes and for training machines to behave like humans.
- Visual Localization using Imperfect 3D Models from the Internet32
A new benchmark for visual localization is created and a detailed experimental evaluation based on multiple 3D models per scene is provided, which shows that3D models from the Internet show promise as an easy-to-obtain scene representation.
- How Privacy-Preserving are Line Clouds? Recovering Scene Details from 3D Lines32
A significant amount of information about the 3D scene geometry is preserved in these line clouds, allowing us to recover the3D point positions and thus to recover image content, and is shown at https://github.com/kunalchelani/Line2Point.
- 31
- 29
- Single-Image Depth Prediction Makes Feature Matching Easier29
It is shown that CNN-based depths inferred from single RGB images are quite helpful, despite their flaws, and allow us to pre-warp images and rectify perspective distortions, to significantly enhance SIFT and BRISK features, enabling more good matches, even when cameras are looking at the same scene but in opposite directions.
- IMAGE-TO-IMAGE TRANSLATION FOR ENHANCED FEATURE MATCHING, IMAGE RETRIEVAL AND VISUAL LOCALIZATION27
This work proposes to use image-to-image translation to transform images from a rendered domain to a captured domain and shows that translated images in the captured domain are of higher quality than the rendered images.
- The Unreasonable Effectiveness of Pre-Trained Features for Camera Pose Refinement26
A simple approach that combines pre-trained features with a particle filter and a renderable representation of the scene and achieves state-of-the-art results, demonstrating that one can easily build a pose refiner without the need for specific training.
- 25
- 25
- Reference Pose Generation for Visual Localization via Learned Features and View Synthesis23
A semi-automated approach to generate reference poses based on feature matching between renderings of a 3D model and real images via learned features is proposed, showing that state-of-the-art visual localization methods perform better than predicted by the original reference poses.
- 23
- Objects Can Move: 3D Change Detection by Geometric Transformation Constistency20
This work proposes a 3D object discovery method that is based only on scene changes and achieves state-of-the-art performance on the 3RScan dataset against competitive baselines.
- CrowdDriven: A New Challenging Dataset for Outdoor Visual Localization20
This work proposes a new benchmark for visual localization in outdoor scenes, using crowd-sourced data to cover a wide range of geographical regions and camera devices with a focus on the failure cases of current algorithms.
- 20
- EdgeGaussians - 3D Edge Mapping via Gaussian Splatting18
This paper proposes a method that explicitly learns 3D edge points with a 3D Gaussian Splatting representation trained from edge images, and produces results better than or at-par with the current state-of-the-art methods, while being an order of magnitude faster.
- The Second Workshop on 3D Reconstruction Meets Semantics: Challenge Results Discussion18
This paper discusses a reconstruction challenge held as a part of the second 3D Reconstruction meets Semantics workshop (3DRMS), featuring both synthetic and real data from outdoor scenes represented by gardens with a variety of bushes, trees, other plants and objects.
- Calibrated and Partially Calibrated Semi-Generalized Homographies16
The first minimal solutions for estimating the semi-generalized homography given a perspective and a generalized camera are proposed, using five 2D-2D image point correspondences induced by a scene plane.
- RePoseD: Efficient Relative Pose Estimation With Known Depth Information12
A novel framework for estimating the relative pose of two cameras from point correspondences with associated monocular depths is proposed, and two new solvers outperform state-of-the-art depth-aware solvers in terms of speed and accuracy.
- P1AC: Revisiting Absolute Pose From a Single Affine Correspondence12
P1AC is introduced, the first general solution to the problem of estimating the pose of a calibrated camera given a single observation of an oriented point and an affine correspondence, and is applicable to large-scale image-based localization.
- Towards Fast Image-Based Localization on a City-Scale12
This paper analyses a recently proposed localization method that achieves state-of-the-art localization performance using a visual vocabulary quantization for efficient 2D-to-3D correspondence search and studies how the size of the subset, as well as the quantization, affect both the search for matches and the time needed by RANSAC for pose estimation.
- A Guide to Structureless Visual Localization11
This paper is dedicated to providing the first comprehensive discussion and comparison of structureless methods, and shows that approaches based on classical absolute or semi-generalized relative pose estimation outperform very recent methods based on pose regression by a wide margin.
- Gaussian Splatting Feature Fields for (Privacy-Preserving) Visual Localization11
3D Gaussian Splatting Feature Fields (GSFFs) is proposed, a scene representation for visual localization that combines an explicit geometry model (3DGS) with an implicit feature field (3DGS) to learn robust feature representations grounded in 3D.
- Privacy-Preserving Representations are not Enough: Recovering Scene Content from Camera Poses10
This paper shows that an attacker can learn about details of a scene without any access by simply querying a localization service, and does not place any requirements on the localization algorithm used, and thus applies to privacy-preserving representations.
- 8
- Comparative Evaluation of 3D Reconstruction Methods for Object Pose Estimation7
This work proposes a novel benchmark for measuring the impact of 3D reconstruction quality on pose estimation accuracy and shows that classical, non-learning-based approaches can perform on par with modern learning-based reconstruction techniques and can even offer a better reconstruction time-pose accuracy tradeoff.
- Obfuscation Based Privacy Preserving Representations Are Recoverable Using Neighborhood Information7
A common vulnerability in geometry obfuscation techniques is revealed that allows approximate point recovery using known neighborhoods, and it is shown that these neighborhoods can be computed by learning to identify descriptors that co-occur in neighborhoods.
- Self-Supervised Learning of Neural Implicit Feature Fields for Camera Pose Refinement7
This paper proposes to jointly learn the scene representation along with a 3D dense feature field and a 2D feature extractor whose outputs are embedded in the same metric space and demonstrates that this approach outperforms prior and concurrent work on leveraging implicit scene representations for localization.
- 7
- Exploiting Spatial and Co-visibility Relations for Image-Based Localization6
This chapter shows that the effectiveness of approaches based on prioritized 2D-to-3D matching at little to no additional run-time costs can be increased by exploiting both spatial and co-visibility relations between the 3D points in the model.
- Has Anything Changed? 3D Change Detection by 2D Segmentation Masks5
An unsupervised object discovery method that identifies added, moved, or removed objects without any prior knowledge of what objects exist in the scene by leveraging generic 2D segmentation masks in the 3D space.
- Partially calibrated semi-generalized pose from hybrid point correspondences4
Practical solvers to previously unsolved challenging configurations are derived and it is shown that in the presence of noise in the 3D points these solvers provide better estimates than the corresponding absolute pose solvers.
- Combining Absolute and Semi-Generalized Relative Poses for Visual Localization3
This work analyzes combining structure-based and structure-less strategies while exploring how to select between poses obtained from 2D-2D and 2D-3D matches, respectively and shows that combining both strategies improves localization performance in multiple practically relevant scenarios.
- Remove360: Benchmarking Residuals After Object Removal in 3D Gaussian Splatting3
A novel benchmark and evaluation framework is introduced to quantify semantic residuals, the unintended cues left behind after object removal in 3D Gaussian Splatting, showing that current methods often preserve semantic information despite the absence of visual geometry.
- 3
- 3
- Practical solutions to the relative pose of three calibrated cameras2
Novel efficient solutions to the challenging problem of estimating the relative pose of three calibrated cameras from four point correspondences are proposed, with the novel solver based on approximate mean-point correspondences being more robust and accurate than the affine-based solver.
- 2
- 2
- Rekonstruktion urbaner Umgebungen und ihre Anwendungen2
Die Erzeugung virtueller Stadtmodelle in hinreichender Qualitat in Bezug auf die verschiedenen Zielapplikationen ist bis heute eine extreme Herausforderung and stellt zumeist eine zeitaufwendige (oft manuelle) Aufgabe dar.
- Benchmarking the Effects of Object Pose Estimation and Reconstruction on Robotic Grasping Success1
A large-scale, physics-based benchmark that evaluates 6D pose estimators and 3D mesh models based on their functional efficacy in grasping provides insight into how perception systems relate to object manipulation using robots.
- Are Minimal Radial Distortion Solvers Really Necessary for Relative Pose Estimation?1
This paper compares radial distortion solvers with two simple-to-implement approaches that do not use minimal radial distortion solvers: the first approach combines an efficient pinhole solver with sampled radial undistortion parameters, where the sampled parameters are used for undistortion prior to applying the pinhole solver.
- Are Minimal Radial Distortion Solvers Necessary for Relative Pose Estimation?1
This paper compares radial distortion solvers with a simple-to-implement approach that combines an efficient pinhole solver with sampled radial distortion parameters and clearly shows that complex radial distortion solvers are not necessary in practice.
- 1
- 1
- Guest Editorial: Special Issue on Performance Evaluation in Computer Vision1
In this special issue, all types of contributions broadly relating to performance evaluation in computer vision are sought, including large, well-designed training datasets and well-defined loss functions.
- 1
- Efficient effective image based localization1
This thesis proposes methods that can localize novel query images taken under a wide range of viewing conditions while requiring only a small amount of processing time and identifies the algorithmic factors preventing image retrieval methods to achieve the same effectiveness as the original system.
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth–
Depth2Pose is proposed, a framework for evaluating MDEs in the context of downstream tasks by combining depth predictions with feature correspondences in depth-aware geometric solvers, and uses relative camera pose estimation accuracy as a task-driven proxy for depth quality.
- DGSfM: Depth-Guided Scale-Aware Global Structure-from-Motion–
DGSfM is proposed, a depth-aware global SfM pipeline that uses monocular depth maps as a scalable prior while preserving explicit multi-view optimization, and consistently improves over strong global SfM baselines across sparse and dense matching front-ends, achieving substantial gains in pose accuracy.
- –
- –
- Privacy-Preserving Structureless Visual Localization via Image Obfuscation–
This paper studies a simple image obfuscation approach based on common image operations, e.g., replacing RGB images with (semantic) segmentations, and shows that existing structureless pipelines do not need any special adjustments, as modern feature matchers can match obfuscated images out of the box.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Large-scale visual SLAM for in-the-wild videos–
This proposed system establishes a new baseline for visual reconstruction from casual uncontrolled videos found online, demonstrating more consistent reconstructions over longer sequences of in-the-wild videos than previously achieved.
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.