Michael Weinmann

Elena Trunz, Jonathan Klein, Jan Müller et al.

We investigate the capabilities of neural inverse procedural modeling to infer high-quality procedural yarn models with fiber-level details from single images of depicted yarn samples. While directly inferring all parameters of the underlying yarn model based on a single neural network may seem an intuitive choice, we show that the complexity of yarn structures in terms of twisting and migration characteristics of the involved fibers can be better encountered in terms of ensembles of networks that focus on individual characteristics. We analyze the effect of different loss functions including a parameter loss to penalize the deviation of inferred parameters to ground truth annotations, a reconstruction loss to enforce similar statistics of the image generated for the estimated parameters in comparison to training images as well as an additional regularization term to explicitly penalize deviations between latent codes of synthetic images and the average latent code of real images in the latent space of the encoder. We demonstrate that the combination of a carefully designed parametric, procedural yarn model with respective network ensembles as well as loss functions even allows robust parameter inference when solely trained on synthetic data. Since our approach relies on the availability of a yarn database with parameter annotations and we are not aware of such a respectively available dataset, we additionally provide, to the best of our knowledge, the first dataset of yarn images with annotations regarding the respective yarn parameters. For this purpose, we use a novel yarn generator that improves the realism of the produced results over previous approaches.

Lukas Bode, Michael Weinmann, Reinhard Klein

Extracting high-level structural information from 3D point clouds is challenging but essential for tasks like urban planning or autonomous driving requiring an advanced understanding of the scene at hand. Existing approaches are still not able to produce high-quality results consistently while being fast enough to be deployed in scenarios requiring interactivity. We propose to utilize a novel set of features describing the local neighborhood on a per-point basis via first and second order statistics as input for a simple and compact classification network to distinguish between non-edge, sharp-edge, and boundary points in the given data. Leveraging this feature embedding enables our algorithm to outperform the state-of-the-art techniques in terms of quality and processing time.

Javier Grau, Markus Plack, Patrick Haehn et al.

Non-line-of-sight reconstruction (NLoS) is a novel indirect imaging modality that aims to recover objects or scene parts outside the field of view from measurements of light that is indirectly scattered off a directly visible, diffuse wall. Despite recent advances in acquisition and reconstruction techniques, the well-posedness of the problem at large, and the recoverability of objects and their shapes in particular, remains an open question. The commonly employed Fermat path criterion is rather conservative with this regard, as it classifies some surfaces as unrecoverable, although they contribute to the signal. In this paper, we use a simpler necessary criterion for an opaque surface patch to be recoverable. Such piece of surface must be directly visible from some point on the wall, and it must occlude the space behind itself. Inspired by recent advances in neural implicit representations, we devise a new representation and reconstruction technique for NLoS scenes that unifies the treatment of recoverability with the reconstruction itself. Our approach, which we validate on various synthetic and experimental datasets, exhibits interesting properties. Unlike memory-inefficient volumetric representations, ours allows to infer adaptively tessellated surfaces from time-of-flight measurements of moderate resolution. It can further recover features beyond the Fermat path criterion, and it is robust to significant amounts of self-occlusion. We believe that this is the first time that these properties have been achieved in one system that, as an additional benefit, is trainable and hence suited for data-driven approaches.

Patrick Stotko, Michael Weinmann, Reinhard Klein

We present incomplete gamma kernels, a generalization of Locally Optimal Projection (LOP) operators. In particular, we reveal the relation of the classical localized $ L_1 $ estimator, used in the LOP operator for point cloud denoising, to the common Mean Shift framework via a novel kernel. Furthermore, we generalize this result to a whole family of kernels that are built upon the incomplete gamma function and each represents a localized $ L_p $ estimator. By deriving various properties of the kernel family concerning distributional, Mean Shift induced, and other aspects such as strict positive definiteness, we obtain a deeper understanding of the operator's projection behavior. From these theoretical insights, we illustrate several applications ranging from an improved Weighted LOP (WLOP) density weighting scheme and a more accurate Continuous LOP (CLOP) kernel approximation to the definition of a novel set of robust loss functions. These incomplete gamma losses include the Gaussian and LOP loss as special cases and can be applied to various tasks including normal filtering. Furthermore, we show that the novel kernels can be included as priors into neural networks. We demonstrate the effects of each application in a range of quantitative and qualitative experiments that highlight the benefits induced by our modifications.

Leif Van Holland, Patrick Stotko, Stefan Krumpen et al.

Despite the impressive progress of telepresence systems for room-scale scenes with static and dynamic scene entities, expanding their capabilities to scenarios with larger dynamic environments beyond a fixed size of a few square-meters remains challenging. In this paper, we aim at sharing 3D live-telepresence experiences in large-scale environments beyond room scale with both static and dynamic scene entities at practical bandwidth requirements only based on light-weight scene capture with a single moving consumer-grade RGB-D camera. To this end, we present a system which is built upon a novel hybrid volumetric scene representation in terms of the combination of a voxel-based scene representation for the static contents, that not only stores the reconstructed surface geometry but also contains information about the object semantics as well as their accumulated dynamic movement over time, and a point-cloud-based representation for dynamic scene parts, where the respective separation from static parts is achieved based on semantic and instance information extracted for the input frames. With an independent yet simultaneous streaming of both static and dynamic content, where we seamlessly integrate potentially moving but currently static scene entities in the static model until they are becoming dynamic again, as well as the fusion of static and dynamic data at the remote client, our system is able to achieve VR-based live-telepresence at close to real-time rates. Our evaluation demonstrates the potential of our novel approach in terms of visual quality, performance, and ablation studies regarding involved design choices.

Saptarshi Neil Sinha, Holger Graf, Michael Weinmann

We propose a novel cross-spectral rendering framework based on 3D Gaussian Splatting (3DGS) that generates realistic and semantically meaningful splats from registered multi-view spectrum and segmentation maps. This extension enhances the representation of scenes with multiple spectra, providing insights into the underlying materials and segmentation. We introduce an improved physically-based rendering approach for Gaussian splats, estimating reflectance and lights per spectra, thereby enhancing accuracy and realism. In a comprehensive quantitative and qualitative evaluation, we demonstrate the superior performance of our approach with respect to other recent learning-based spectral scene representation approaches (i.e., XNeRF and SpectralNeRF) as well as other non-spectral state-of-the-art learning-based approaches. Our work also demonstrates the potential of spectral scene understanding for precise scene editing techniques like style transfer, inpainting, and removal. Thereby, our contributions address challenges in multi-spectral scene representation, rendering, and editing, offering new possibilities for diverse applications.

Andre Rochow, Max Schwarz, Michael Weinmann et al.

Novel view synthesis is required in many robotic applications, such as VR teleoperation and scene reconstruction. Existing methods are often too slow for these contexts, cannot handle dynamic scenes, and are limited by their explicit depth estimation stage, where incorrect depth predictions can lead to large projection errors. Our proposed method runs in real time on live streaming data and avoids explicit depth estimation by efficiently warping input images into the target frame for a range of assumed depth planes. The resulting plane sweep volume (PSV) is directly fed into our network, which first estimates soft PSV masks in a self-supervised manner, and then directly produces the novel output view. This improves efficiency and performance on transparent, reflective, thin, and feature-less scene parts. FaDIV-Syn can perform both interpolation and extrapolation tasks at 540p in real-time and outperforms state-of-the-art extrapolation methods on the large-scale RealEstate10k dataset. We thoroughly evaluate ablations, such as removing the Soft-Masking network, training from fewer examples as well as generalization to higher resolutions and stronger depth discretization. Our implementation is available.

Dennis Haitz, Athradi Shritish Shetty, Michael Weinmann et al.

Visual Place Recognition (VPR) is a core component in computer vision, typically formulated as an image retrieval task for localization, mapping, and navigation. In this work, we instead study VPR as an image pair retrieval front-end for registration pipelines, where the goal is to find top-matching image pairs between two disjoint image sets for downstream tasks such as scene registration, SLAM, and Structure-from-Motion. We comparatively evaluate state-of-the-art VPR families - NetVLAD-style baselines, classification-based global descriptors (CosPlace, EigenPlaces), feature-mixing (MixVPR), and foundation-model-driven methods (AnyLoc, SALAD, MegaLoc) - on three challenging datasets: object-centric outdoor scenes (Tanks and Temples), indoor RGB-D scans (ScanNet-GS), and autonomous-driving sequences (KITTI). We show that modern global descriptor approaches are increasingly suitable as off-the-shelf image pair retrieval modules in challenging scenarios including perceptual aliasing and incomplete sequences, while exhibiting clear, domain-dependent strengths and weaknesses that are critical when choosing VPR components for robust mapping and registration.

Saptarshi Neil Sinha, Julius Kühn, Mika Silvan Goschke et al.

Automated and selective harvesting of fruits has become an important area of research, particularly due to challenges such as high costs and a shortage of seasonal labor in advanced economies. This paper focuses on 6D pose estimation of strawberries using purely synthetic data generated through a procedural pipeline for photorealistic rendering. We employ the YOLOX-6D-Pose algorithm, a single-shot approach that leverages the YOLOX backbone, known for its balance between speed and accuracy, and its support for edge inference. To address the lacking availability of training data, we introduce a robust and flexible pipeline for generating synthetic strawberry data from various 3D models via a procedural Blender pipeline, where we focus on enhancing the realism of the synthesized data in comparison to previous work to make it a valuable resource for training pose estimation algorithms. Quantitative evaluations indicate that our models achieve comparable accuracy on both the NVIDIA RTX 3090 and Jetson Orin Nano across several ADD-S metrics, with the RTX 3090 demonstrating superior processing speed. However, the Jetson Orin Nano is particularly suited for resource-constrained environments, making it an excellent choice for deployment in agricultural robotics. Qualitative assessments further confirm the model's performance, demonstrating its capability to accurately infer the poses of ripe and partially ripe strawberries, while facing challenges in detecting unripe specimens. This suggests opportunities for future improvements, especially in enhancing detection capabilities for unripe strawberries (if desired) by exploring variations in color. Furthermore, the methodology presented could be adapted easily for other fruits such as apples, peaches, and plums, thereby expanding its applicability and impact in the field of agricultural automation.

Leif Van Holland, Michael Weinmann, Jan U. Müller et al.

While neural radiance fields (NeRF) led to a breakthrough in photorealistic novel view synthesis, handling mirroring surfaces still denotes a particular challenge as they introduce severe inconsistencies in the scene representation. Previous attempts either focus on reconstructing single reflective objects or rely on strong supervision guidance in terms of additional user-provided annotations of visible image regions of the mirrors, thereby limiting the practical usability. In contrast, in this paper, we present NeRF-MD, a method which shows that NeRFs can be considered as mirror detectors and which is capable of reconstructing neural radiance fields of scenes containing mirroring surfaces without the need for prior annotations. To this end, we first compute an initial estimate of the scene geometry by training a standard NeRF using a depth reprojection loss. Our key insight lies in the fact that parts of the scene corresponding to a mirroring surface will still exhibit a significant photometric inconsistency, whereas the remaining parts are already reconstructed in a plausible manner. This allows us to detect mirror surfaces by fitting geometric primitives to such inconsistent regions in this initial stage of the training. Using this information, we then jointly optimize the radiance field and mirror geometry in a second training stage to refine their quality. We demonstrate the capability of our method to allow the faithful detection of mirrors in the scene as well as the reconstruction of a single consistent scene representation, and demonstrate its potential in comparison to baseline and mirror-aware approaches.

Benno Buschmann, Andreea Dogaru, Elmar Eisemann et al.

Learning-based scene representations such as neural radiance fields or light field networks, that rely on fitting a scene model to image observations, commonly encounter challenges in the presence of inconsistencies within the images caused by occlusions, inaccurately estimated camera parameters or effects like lens flare. To address this challenge, we introduce RANdom RAy Consensus (RANRAC), an efficient approach to eliminate the effect of inconsistent data, thereby taking inspiration from classical RANSAC based outlier detection for model fitting. In contrast to the down-weighting of the effect of outliers based on robust loss formulations, our approach reliably detects and excludes inconsistent perspectives, resulting in clean images without floating artifacts. For this purpose, we formulate a fuzzy adaption of the RANSAC paradigm, enabling its application to large scale models. We interpret the minimal number of samples to determine the model parameters as a tunable hyperparameter, investigate the generation of hypotheses with data-driven models, and analyze the validation of hypotheses in noisy environments. We demonstrate the compatibility and potential of our solution for both photo-realistic robust multi-view reconstruction from real-world images based on neural radiance fields and for single-shot reconstruction based on light-field networks. In particular, the results indicate significant improvements compared to state-of-the-art robust methods for novel-view synthesis on both synthetic and captured scenes with various inconsistencies including occlusions, noisy camera pose estimates, and unfocused perspectives. The results further indicate significant improvements for single-shot reconstruction from occluded images. Project Page: https://bennobuschmann.com/ranrac/

Paul Julius Kühn, Cedric Spengler, Michael Weinmann et al.

Image-based shape retrieval (IBSR) aims to retrieve 3D models from a database given a query image, hence addressing a classical task in computer vision, computer graphics, and robotics. Recent approaches typically rely on bridging the domain gap between 2D images and 3D shapes based on the use of multi-view renderings as well as task-specific metric learning to embed shapes and images into a common latent space. In contrast, we address IBSR through large-scale multi-modal pretraining and show that explicit view-based supervision is not required. Inspired by pre-aligned image--point-cloud encoders from ULIP and OpenShape that have been used for tasks such as 3D shape classification, we propose the use of pre-aligned image and shape encoders for zero-shot and standard IBSR by embedding images and point clouds into a shared representation space and performing retrieval via similarity search over compact single-embedding shape descriptors. This formulation allows skipping view synthesis and naturally enables zero-shot and cross-domain retrieval without retraining on the target database. We evaluate pre-aligned encoders in both zero-shot and supervised IBSR settings and additionally introduce a multi-modal hard contrastive loss (HCL) to further increase retrieval performance. Our evaluation demonstrates state-of-the-art performance, outperforming related methods on $Acc_{Top1}$ and $Acc_{Top10}$ for shape retrieval across multiple datasets, with best results observed for OpenShape combined with Point-BERT. Furthermore, training on our proposed multi-modal HCL yields dataset-dependent gains in standard instance retrieval tasks on shape-centric data, underscoring the value of pretraining and hard contrastive learning for 3D shape retrieval. The code will be made available via the project website.

Saptarshi Neil Sinha, P. Julius Kuehn, Johannes Koppe et al.

The preservation of early visual arts, particularly color photographs, is challenged by deterioration caused by aging and improper storage, leading to issues like blurring, scratches, color bleeding, and fading defects. Despite great advances in image restoration and enhancement in recent years, such systematic defects often cannot be restored by current state-of-the-art software features as available e.g. in Adobe Photoshop, but would require the incorporation of defect-aware priors into the underlying machine learning techniques. However, there are no publicly available datasets of autochromes with defect annotations. In this paper, we address these limitations and present the first approach that allows the automatic removal of greening color defects in digitized autochrome photographs. For this purpose, we introduce an approach for accurately simulating respective defects and use the respectively obtained synthesized data with its ground truth defect annotations to train a generative AI model with a carefully designed loss function that accounts for color imbalances between defected and non-defected areas. As demonstrated in our evaluation, our approach allows for the efficient and effective restoration of the considered defects, thereby overcoming limitations of alternative techniques that struggle with accurately reproducing original colors and may require significant manual effort.

Nils Wandel, Michael Weinmann, Michael Neidlin et al.

Partial Differential Equations (PDEs) are notoriously difficult to solve. In general, closed-form solutions are not available and numerical approximation schemes are computationally expensive. In this paper, we propose to approach the solution of PDEs based on a novel technique that combines the advantages of two recently emerging machine learning based approaches. First, physics-informed neural networks (PINNs) learn continuous solutions of PDEs and can be trained with little to no ground truth data. However, PINNs do not generalize well to unseen domains. Second, convolutional neural networks provide fast inference and generalize but either require large amounts of training data or a physics-constrained loss based on finite differences that can lead to inaccuracies and discretization artifacts. We leverage the advantages of both of these approaches by using Hermite spline kernels in order to continuously interpolate a grid-based state representation that can be handled by a CNN. This allows for training without any precomputed training data using a physics-informed loss function only and provides fast, continuous solutions that generalize to unseen domains. We demonstrate the potential of our method at the examples of the incompressible Navier-Stokes equation and the damped wave equation. Our models are able to learn several intriguing phenomena such as Karman vortex streets, the Magnus effect, Doppler effect, interference patterns and wave reflections. Our quantitative assessment and an interactive real-time demo show that we are narrowing the gap in accuracy of unsupervised ML based methods to industrial CFD solvers while being orders of magnitude faster.

Stefan Krumpen, Reinhard Klein, Michael Weinmann

VR/AR technology is a key enabler for new ways of immersively experiencing cultural heritage artifacts based on their virtual counterparts obtained from a digitization process. In this paper, we focus on enriching VR-based object inspection by additional haptic feedback, thereby creating tangible cultural heritage experiences. For this purpose, we present an approach for interactive and collaborative VR-based object inspection and annotation. Our system supports high-quality 3D models with accurate reflectance characteristics while additionally providing haptic feedback regarding the object shape features based on a 3D printed replica. The digital object model in terms of a printable representation of the geometry as well as reflectance characteristics are stored in a compact and streamable representation on a central server, which streams the data to remotely connected users/clients. The latter can jointly perform an interactive inspection of the object in VR with additional haptic feedback through the 3D printed replica. Evaluations regarding system performance, visual quality of the considered models as well as insights from a user study indicate an improved interaction, assessment and experience of the considered objects.

Nils Wandel, Michael Weinmann, Reinhard Klein

Physically plausible fluid simulations play an important role in modern computer graphics and engineering. However, in order to achieve real-time performance, computational speed needs to be traded-off with physical accuracy. Surrogate fluid models based on neural networks have the potential to achieve both, fast fluid simulations and high physical accuracy. However, these approaches rely on massive amounts of training data, require complex pipelines for training and inference or do not generalize to new fluid domains. In this work, we present significant extensions to a recently proposed deep learning framework, which addresses the aforementioned challenges in 2D. We go from 2D to 3D and propose an efficient architecture to cope with the high demands of 3D grids in terms of memory and computational complexity. Furthermore, we condition the neural fluid model on additional information about the fluid's viscosity and density which allows simulating laminar as well as turbulent flows based on the same surrogate model. Our method allows to train fluid models without requiring fluid simulation data beforehand. Inference is fast and simple, as the fluid model directly maps a fluid state and boundary conditions at a moment t to a subsequent fluid state at t+dt. We obtain real-time fluid simulations on a 128x64x64 grid that include various fluid phenomena such as the Magnus effect or Karman vortex streets and generalize to domain geometries not considered during training. Our method indicates strong improvements in terms of accuracy, speed and generalization capabilities over current 3D NN-based fluid models.

Nils Wandel, Michael Weinmann, Reinhard Klein

Fast and stable fluid simulations are an essential prerequisite for applications ranging from computer-generated imagery to computer-aided design in research and development. However, solving the partial differential equations of incompressible fluids is a challenging task and traditional numerical approximation schemes come at high computational costs. Recent deep learning based approaches promise vast speed-ups but do not generalize to new fluid domains, require fluid simulation data for training, or rely on complex pipelines that outsource major parts of the fluid simulation to traditional methods. In this work, we propose a novel physics-constrained training approach that generalizes to new fluid domains, requires no fluid simulation data, and allows convolutional neural networks to map a fluid state from time-point t to a subsequent state at time t + dt in a single forward pass. This simplifies the pipeline to train and evaluate neural fluid models. After training, the framework yields models that are capable of fast fluid simulations and can handle various fluid phenomena including the Magnus effect and Karman vortex streets. We present an interactive real-time demo to show the speed and generalization capabilities of our trained models. Moreover, the trained neural networks are efficient differentiable fluid solvers as they offer a differentiable update step to advance the fluid simulation in time. We exploit this fact in a proof-of-concept optimal control experiment. Our models significantly outperform a recent differentiable fluid solver in terms of computational speed and accuracy.

Julian Tanke, Oh-Hun Kwon, Patrick Stotko et al.

The key prerequisite for accessing the huge potential of current machine learning techniques is the availability of large databases that capture the complex relations of interest. Previous datasets are focused on either 3D scene representations with semantic information, tracking of multiple persons and recognition of their actions, or activity recognition of a single person in captured 3D environments. We present Bonn Activity Maps, a large-scale dataset for human tracking, activity recognition and anticipation of multiple persons. Our dataset comprises four different scenes that have been recorded by time-synchronized cameras each only capturing the scene partially, the reconstructed 3D models with semantic annotations, motion trajectories for individual people including 3D human poses as well as human activity annotations. We utilize the annotations to generate activity likelihoods on the 3D models called activity maps.

Jan Müller, Reinhard Klein, Michael Weinmann

Wasserstein-GANs have been introduced to address the deficiencies of generative adversarial networks (GANs) regarding the problems of vanishing gradients and mode collapse during the training, leading to improved convergence behaviour and improved image quality. However, Wasserstein-GANs require the discriminator to be Lipschitz continuous. In current state-of-the-art Wasserstein-GANs this constraint is enforced via gradient norm regularization. In this paper, we demonstrate that this regularization does not encourage a broad distribution of spectral-values in the discriminator weights, hence resulting in less fidelity in the learned distribution. We therefore investigate the possibility of substituting this Lipschitz constraint with an orthogonality constraint on the weight matrices. We compare three different weight orthogonalization techniques with regards to their convergence properties, their ability to ensure the Lipschitz condition and the achieved quality of the learned distribution. In addition, we provide a comparison to Wasserstein-GANs trained with current state-of-the-art methods, where we demonstrate the potential of solely using orthogonality-based regularization. In this context, we propose an improved training procedure for Wasserstein-GANs which utilizes orthogonalization to further increase its generalization capability. Finally, we provide a novel metric to evaluate the generalization capabilities of the discriminators of different Wasserstein-GANs.

Patrick Stotko, Stefan Krumpen, Michael Weinmann et al.

Sharing live telepresence experiences for teleconferencing or remote collaboration receives increasing interest with the recent progress in capturing and AR/VR technology. Whereas impressive telepresence systems have been proposed on top of on-the-fly scene capture, data transmission and visualization, these systems are restricted to the immersion of single or up to a low number of users into the respective scenarios. In this paper, we direct our attention on immersing significantly larger groups of people into live-captured scenes as required in education, entertainment or collaboration scenarios. For this purpose, rather than abandoning previous approaches, we present a range of optimizations of the involved reconstruction and streaming components that allow the immersion of a group of more than 24 users within the same scene - which is about a factor of 6 higher than in previous work - without introducing further latency or changing the involved consumer hardware setup. We demonstrate that our optimized system is capable of generating high-quality scene reconstructions as well as providing an immersive viewing experience to a large group of people within these live-captured scenes.

Patrick Stotko, Stefan Krumpen, Max Schwarz et al.

Applications like disaster management and industrial inspection often require experts to enter contaminated places. To circumvent the need for physical presence, it is desirable to generate a fully immersive individual live teleoperation experience. However, standard video-based approaches suffer from a limited degree of immersion and situation awareness due to the restriction to the camera view, which impacts the navigation. In this paper, we present a novel VR-based practical system for immersive robot teleoperation and scene exploration. While being operated through the scene, a robot captures RGB-D data that is streamed to a SLAM-based live multi-client telepresence system. Here, a global 3D model of the already captured scene parts is reconstructed and streamed to the individual remote user clients where the rendering for e.g. head-mounted display devices (HMDs) is performed. We introduce a novel lightweight robot client component which transmits robot-specific data and enables a quick integration into existing robotic systems. This way, in contrast to first-person exploration systems, the operators can explore and navigate in the remote site completely independent of the current position and view of the capturing robot, complementing traditional input devices for teleoperation. We provide a proof-of-concept implementation and demonstrate the capabilities as well as the performance of our system regarding interactive object measurements and bandwidth-efficient data streaming and visualization. Furthermore, we show its benefits over purely video-based teleoperation in a user study revealing a higher degree of situation awareness and a more precise navigation in challenging environments.

Rodrigo Martín, Michael Weinmann, Matthias B. Hullin

Even in the digital age, designers largely rely on physical material samples to illustrate their products, as existing visual representations fail to sufficiently reproduce the look and feel of real world materials. Here, we investigate the use of interactive material sonification as an additional sensory modality for communicating well-established material qualities like softness, pleasantness or value. We developed a custom application for touchscreen devices that receives tactile input and translate it into material rubbing sound using granular synthesis. We used this system to perform a psychophysical study, in which the ability of the user to rate subjective material qualities is evaluated, with the actual material samples serving as reference stimulus. Our experimental results indicate that the considered audio cues do not significantly contribute to the perception of material qualities but are able to increase the level of immersion when interacting with digital samples.

Patrick Stotko, Stefan Krumpen, Matthias B. Hullin et al.

Real-time 3D scene reconstruction from RGB-D sensor data, as well as the exploration of such data in VR/AR settings, has seen tremendous progress in recent years. The combination of both these components into telepresence systems, however, comes with significant technical challenges. All approaches proposed so far are extremely demanding on input and output devices, compute resources and transmission bandwidth, and they do not reach the level of immediacy required for applications such as remote collaboration. Here, we introduce what we believe is the first practical client-server system for real-time capture and many-user exploration of static 3D scenes. Our system is based on the observation that interactive frame rates are sufficient for capturing and reconstruction, and real-time performance is only required on the client site to achieve lag-free view updates when rendering the 3D model. Starting from this insight, we extend previous voxel block hashing frameworks by overcoming internal dependencies and introducing, to the best of our knowledge, the first thread-safe GPU hash map data structure that is robust under massively concurrent retrieval, insertion and removal of entries on a thread level. We further propose a novel transmission scheme for volume data that is specifically targeted to Marching Cubes geometry reconstruction and enables a 90% reduction in bandwidth between server and exploration clients. The resulting system poses very moderate requirements on network bandwidth, latency and client-side computation, which enables it to rely entirely on consumer-grade hardware, including mobile devices. We demonstrate that our technique achieves state-of-the-art representation accuracy while providing, for any number of clients, an immersive and fluid lag-free viewing experience even during network outages.