As we prepare for our demonstrator at MWC 2026, we’ve been creating high-fidelity 3D Gaussian Splat scenes to serve as immersive background stages for our holographic telepresence experience. Among the various approaches we’ve explored, two stand out for their versatility, flexibility, and ease of use. In this side-by-side comparison: 🔴 Left: Nephelie Technologies Limited’s own rendering, generated using our wireless volumetric video capture system 🔵 Right: The XGRIDS PortalCam, the market leader in popular handheld scanners combining LiDAR and fisheye cameras Both approaches produce detailed, photorealistic spatial environments that can serve as digital stages behind live volumetric captures. Our next step is to combine the strengths of each method to move toward the most versatile and highest-quality 3D scanning pipeline for Gaussian Splat rendering. So, stay tuned! If you’re attending MWC 2026, come and meet us at the JOINER Platform stand — Hall 7, Booth 7B47 — to see the results firsthand and experience the future of holographic telepresence. #6G #TelecomInnovation #MWC2026 #DigitalTwins #Connectivity #FutureOfTelecom #JOINER #3DGS #4DGS #VolumetricVideo #GaussianSplatting #VirtualProduction #XR #VFX #3DScanning #RadianceFields #spatialfields #CreativeTools #3DWeb

𝗙𝗿𝗼𝗺 𝗥𝗲𝗮𝗹 𝗧𝗲𝗿𝗿𝗮𝗶𝗻 𝘁𝗼 𝗮 𝗦𝗶𝗺𝘂𝗹𝗮𝘁𝗶𝗼𝗻 𝗘𝗻𝘃𝗶𝗿𝗼𝗻𝗺𝗲𝗻𝘁 𝗶𝗻 𝟱 𝗛𝗼𝘂𝗿𝘀 Most simulation environments are not built from real terrain. They are usually manually modeled from DEMs, orthophotos, maps, or simplified 3D assets. As a result, many simulations operate on terrain that approximates reality rather than reproducing it. So, what happens when the environment is generated directly from the terrain itself? This animation shows a digital twin generated from aerial photogrammetry in the Tagus River valley between the Castillo de Oreja and the Despoblado de Oreja in Toledo, Spain. The goal of this experiment was simple. Measure how quickly a real landscape can be converted into a simulation ready virtual environment. The workflow combines aerial capture using the drone DJI Matrice 4E and photogrammetric reconstruction optimized for real time simulation environments such as Unreal Engine. Key metrics from this test area • Surface digitized 5 hectares • Images captured 3,500 • Flight altitude approximately 40 m • Flight operations about 1.5 batteries • Total flight time 1 hour 15 minutes • Original reconstruction 266M triangles • Simulation model 300K triangles • Photogrammetric reconstruction time 3 hours • Processing workstation RTX 5080 GPU and Intel i9 CPU • Environment preparation for simulation 1 hour • Total pipeline time from capture to simulation environment about 5 hours The key point is not just terrain mapping but how quickly real locations can become operational digital environments for simulation, training, or territorial analysis. Possible applications include simulation of mobility in complex terrain, virtual training environments derived from real landscapes, testing and validation of autonomous systems, emergency planning, and infrastructure security analysis. Traditional comparison Simulation environments are often constructed manually from heterogeneous geospatial datasets and manual terrain modeling. Even for relatively small areas this process can take several days depending on the required detail. In this experiment the terrain geometry is derived directly from aerial photogrammetry, allowing a real environment to become simulation ready in approximately five hours. Scalability This test focuses on a 5-hectare terrain segment to measure pipeline speed. However, the same workflow scales to larger areas. Dronícola has previously processed aerial photogrammetry datasets covering up to approximately 200 hectares of terrain. Larger areas increase image counts and flight missions, while reconstruction can be divided into spatial tiles and optimized into modular terrain segments for real time simulation. If generating simulation environments directly from real terrain is becoming this fast, should manual terrain modeling still be the default approach? #DigitalTwins #Photogrammetry #Simulation #DroneMapping #Geomatics  

Leica Geosystems part of Hexagon is bringing the Reality Capture User Conference back to Miami Beach this June 8-11th, 2026 - hosted at the Nobu Hotel. By popular demand, the Laser Scanning User Conference (formerly HDS UC) returns with a strong focus on real-world workflows, user-led case studies, training sessions, and open discussions across LiDAR, 3D scanning, digital twins, AEC, surveying, and more. Expect practical insights, peer networking, and conversations that push the reality capture industry forward. More details & registration ↓ https://lnkd.in/ePZ-DHmt #laserscanning #digitaltwin #userconference #casestudies #digitaltwins

In 2008, Radiohead released House of Cards — one of the first mainstream music videos created entirely from 3D scan data using LiDAR and structured light. What made it even more remarkable was that they released the raw 3D data publicly, inviting developers, artists, and technologists to experiment with, remix, and reinterpret it. It was an early open-data moment for spatial media. Since then, more than a dozen volumetric capture technologies have evolved — from depth cameras and photogrammetry to LiDAR, industrial CT, NeRFs, Gaussian splats, and volumetric video stages. The Voxon Photonics 3D Volumetric VLED Platform enables that volumetric data to be replayed in the same form in which it was captured — no glasses, no headsets, unlimited viewing angles, and every point of light emitted from a unique position in space. It’s data you can walk around and view from any direction, even from above. To find out more, please visit www.voxon.co or email contact@voxon.co . The 3D data used is available here: https://lnkd.in/g6YyTwUX

Back from Industry Day at Ft. Irwin, it was clear how critical high-fidelity 3D understanding is for every mission. Emesent’s GX1 SLAM LiDAR shows what’s possible: rapid, high-accuracy 3D capture with integrated 360° imagery that gives teams the detail they need to plan, rehearse, and operate with confidence. Now on the 𝐃𝐞𝐯𝐗 𝐀𝐮𝐭𝐨𝐧𝐨𝐦𝐲  platform, with their solution officially designated 𝐚𝐰𝐚𝐫𝐝-𝐫𝐞𝐚𝐝𝐲! Want to see the level of detail for yourself? Download the GX1 dataset and explore the scan firsthand: 🔗 https://lnkd.in/gvWa2VVi #GX1 #SLAM #LiDAR #MobileMapping #DevXAutonomy #DefenseTech #NTC #FortIrwin #Emesent

🚶 Giving egocentric video models a sense of space — and making them safer. "Spatial-Conditioned Reasoning in Long-Egocentric Videos" studies how depth information improves VLM understanding of first-person video. Key contributions: → Sanpo-D: refined annotations of Google's Sanpo egocentric navigation dataset → Integrates depth data with RGB footage for spatially-grounded representations → Tests multiple VLMs on navigation-focused spatial reasoning tasks → Depth integration improves detection of pedestrians and obstacles (safety-critical) → Trade-off identified: spatial specialization vs. broader accuracy Egocentric video understanding matters for AR/VR, assistive robotics, and first-person activity recognition. But current VLMs often struggle with "where is that object relative to me?" — a fundamental navigation question. Grounding vision in 3D space, not just 2D pixels, is the right direction. #EgocentricAI #VideoUnderstanding #SpatialReasoning #VLM #ARNavigation

🏔 Reconstructing Milano Cortina 2026: Going for Gold in Multi-View Geometry The 2026 Winter Olympics have showcased how breakthroughs in computer vision are happening at 80 mph on a downhill slope and not only in conference halls. Broadcasters deployed around 800+ cameras and 1,800 microphones, capturing over 6,500 hours of content streamed through 44 simultaneous UHD feeds, which is an order of magnitude beyond traditional broadcast infrastructure.  📸 AI Curling Tracking (Real-Time Trajectory Estimation) Curling sheets in Cortina are now equipped with a 12-camera array, treating ice like a laboratory. At ~60 FPS per camera (~720 frames/sec fused), the system reconstructs stone trajectory, velocity vectors, and crucially rotational spin in real-time. It's not just tracking; it is physics-aware reconstruction at broadcast latency. 🔁 360° AI Replays & Volumetric Reconstruction Arrays of 20–50 synchronized cameras capture performances from every angle, enabling 360-degree replays of figure skating jumps and multi-athlete ski cross finishes by fusing hundreds of viewpoints into a single coherent 3D scene. At its core, this is real-time multi-view reconstruction: foreground segmentation, triangulation, and neural view synthesis combined into a latency-constrained pipeline inspired by NeRF-style neural rendering, all running at broadcast scale. 🚁 FPV Drones & Moving-Camera Reconstruction First to transition from static broadcasting rigs to a high-velocity, mobile sensor network using roughly 25-30 drones, including around 15 dedicated FPV units deployed across various venues. Essentially shifting the challenge from static multi-view geometry to dynamic reconstruction with changing extrinsics. Solving it requires real-time camera pose estimation, motion blur compensation, and visual tracking under high acceleration, maintaining spatial consistency despite non-linear flight paths. In essence, it’s SLAM-grade reasoning adapted for live broadcast. There is no longer just filming sports, but reconstructing them in higher resolution with physics-aware 3D intelligence at scale. #ComputerVision #MilanoCortina2026 #3DReconstruction #MultimodalAI #NeuralRendering #SLAM #WinterOlympics2026

When disaster strikes whether from an adversarial event, a natural disaster, or even during routine operations, time and situational awareness are critical. What if you could simply deploy a drone, capture a set of images, and build a detailed 3D map of the environment in minutes? With the latest capabilities from Unknot.id, this is becoming a reality. In this example, we generated a fully interactive 3D representation in just 3 minutes using only 6.25% of the available imagery. The result is a navigable environment where you can zoom, explore, and analyze the scene in real time. 👉 Try the interactive scene: https://lnkd.in/e9gJ_U9J Even more powerful, this capability can run entirely on-device on Android (edge) or scale to cloud/server deployments depending on mission needs. While Unknot.id initially focused on precise 3D positioning, we are now expanding toward context-aware situational intelligence, enabling what we call 4D Location Intelligence. This means not just knowing where things are, but also understanding: • What has changed • Where potential threats may exist • What damage has occurred • How safe an area is for responders or operators The goal is simple: turn raw imagery into actionable spatial intelligence in real time. #AI #ComputerVision #DroneTechnology #3DMapping #SpatialComputing #Geospatial #PublicSafety #DefenseTech #SituationalAwareness #EdgeAI

A serious feature...a not so serious test 😅 Dynamic occlusion is what makes virtual objects appear properly hidden by real-world objects — so your AR model goes behind a tree, fence, terrain, or building instead of floating awkwardly in front of everything. Using camera depth sensors and scene recognition, it enables realistic interactions and enhances immersion by correcting depth perception. Occlusion is one of those “you only notice it when it’s missing” features: without it: models look floaty / wrong with it: the brain immediately accepts the scene as real It’s a small technical detail that delivers a big leap in realism — and helps stakeholders to better understanding visual impact. 🤖➡️🌳❓ #AugmentedReality #AR #SpatialComputing #DigitalTwins #Infrastructure #VizAR #FieldTech #3DVisualization #MixedReality

Over the past months, early access to DroneTwins360 has generated actionable insights from real operational datasets: • 600+ processed datasets • 35,000+ drone and camera images validated through the platform Each dataset, each workflow, and every piece of feedback strengthens intake-stage validation, verified 2D/3D reconstruction, and defensible digital twin generation. A special thank you to early adopters who engaged with the platform. Your trust enables innovation at scale and directly informs enterprise workflows for property claims, inspections, and compliance-driven operations. We are still early in the journey, but thanks to you, it is becoming real. More refinements, experiments, and scalable deployments are coming soon. Visual Overview: • Upload → TrustRank & Preview → Modelling → Export • Dataset verification: metadata, geo-plausibility, completeness • Integration-ready outputs: PLY / Orthomosaic models, BIM, digital twins #DroneTwins360 #UAVData #DigitalTwins #Photogrammetry #DataIntegrity #AviationTech #EnterpriseInnovation #DecisionGrade #DroneSurveying #TrustRank