3d Design Techniques

Big news for the 3D computer vision community! 🙌 ByteDance released Depth Anything 3 on Hugging Face 🔥. This is the world's most powerful model for 3D understanding: it predicts spatially consistent geometry (depth and ray maps) from an arbitrary number of visual inputs, with or without known camera poses. In other words, it allows you to reconstruct a 3D scene just from 2D inputs. DA3 extends monocular depth estimation to any-view scenarios, hence the model can take in single images, multi-view images, and video. Interestingly, the authors reveal two key insights: - A plain transformer (e.g., vanilla DINO) is enough. No specialized architecture is required. - A single depth-ray representation objective is enough. The model does not require a complex multi-task training. Three series of models have been released: the main DA3 series, a monocular metric estimation series, and a monocular depth estimation series. Metric estimation, also called absolute estimation, determines the distance in meters relative to the camera, whereas monocular depth estimation determines the distance relative among the pixels. The authors also released a new visual geometry benchmark covering camera pose estimation, any-view geometry and visual rendering. DA3 sets a new state-of-the-art across all 10 tasks, surpassing prior SOTA, Meta's VGGT, by an average of 35.7% in camera pose accuracy and 23.6% in geometric accuracy. Furthermore, DA3 facilitates SLAM (Simultaneous Localization and Mapping) and 3D Gaussian Splatting by providing a robust and generalizable method for predicting spatially consistent geometry from various visual inputs. Links: - Models: https://lnkd.in/eFFHJhJx - Paper: https://lnkd.in/ewtxy7p6 - Demo: https://lnkd.in/e7Qr3tnG - Code: https://lnkd.in/e89B6JpR

🚀 Fʀᴏᴍ Pᴏɪɴᴛ Cʟᴏᴜᴅꜱ ᴛᴏ Pʀᴇᴄɪꜱɪᴏɴ: Cᴏʀʀɪᴅᴏʀ Mᴀᴘᴘɪɴɢ ᴡɪᴛʜ AʀᴄGIS Pʀᴏ 🌐 Ever wondered how point clouds can be transformed into actionable insights for corridor analytics? Let me take you through a workflow starting from an unclassified Point Cloud! 🛰️✨ Besides using out-of-the-box tools for ground and building classification you also can use deep learning models to classify point clouds. In this case, LiDAR point clouds were additionally classified into Powerlines using a 𝐩𝐫𝐞-𝐭𝐫𝐚𝐢𝐧𝐞𝐝 𝐃𝐞𝐞𝐩 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐌𝐨𝐝𝐞𝐥 𝐟𝐫𝐨𝐦 𝐭𝐡𝐞 𝐄𝐬𝐫𝐢 𝐋𝐢𝐯𝐢𝐧𝐠 𝐀𝐭𝐥𝐚𝐬. The classified Point Cloud then was used to derive raster information such as elevation layers and also features, such as line features for the powerlines (Have a look at the complete workflow in the Storymap (LINK in Comments). This paved the way for leveraging the 3D Analyst tools in ArcGIS Pro. From corridor mapping to spatial analytics, ArcGIS Pro became the hub where data turned into decisions! 🌟 This approach not only simplifies the complexities of point cloud data but also enables accurate and efficient corridor mapping planning, environmental, crucial for infrastructure studies, and beyond. 🛤️🌿 💡 By combining AI with advanced GIS tools, we’re redefining the way we analyze and visualize our world in 3D. 🧠📊 Have a look at the Storymap for the complete workflow! Link in the Comments! 🤝 Let's spark a conversation! How are you leveraging Point Clouds in ArcGIS? Share your insights, challenges, and success stories below. Let's amplify our collective GIS capabilities! 💬💡 💡 #IMGIS25 #LandCover #3DAnalytics #Geospatial #LiDAR #Innovation #CorridorMapping #DeepLearning #Sustainability #Forests #ArcGISPro #Esri #GIS #SpatialAnalysis #ArcGIS #remotesensing #EsriDeutschland #EarthObservation #EsriLivingAtlas #DataVisualization #esrivoices🔍 🚀 🌱

Without 𝐁𝐈𝐌 it was impossible to bring Paris’ Notre-Dame Cathedral to life again 👏 . After the 2019 𝐟𝐢𝐫𝐞 that nearly destroyed the 𝟖𝟔𝟎-year-old landmark Paris’ Notre-Dame Cathedral has been reconstructed .  The rebuilding of Notre Dame was made possible through exceptional craftsmanship, nearly $𝟏 𝐛𝐢𝐥𝐥𝐢𝐨𝐧 in donations from around the world, and a shared resolve to restore the historic site to its former glory. 𝐓𝐡𝐚𝐧𝐤𝐬 𝐭𝐨 𝐀𝐮𝐭𝐨𝐝𝐞𝐬𝐤  𝐂𝐀𝐃 software firms were in full swing offering to provide their AEC tools to the project 𝐟𝐨𝐫 𝐟𝐫𝐞𝐞. 𝐀𝐮𝐭𝐨𝐝𝐞𝐬𝐤 provided some licenses of its AEC suites, with access to 𝖠𝗎𝗍𝗈𝖢𝖠𝖣, 𝖱𝖾𝖼𝖺𝗉 (𝗅𝖺𝗌𝖾𝗋 𝗌𝖼𝖺𝗇𝗇𝗂𝗇𝗀) 𝖺𝗇𝖽 𝖱𝖾𝗏𝗂𝗍 𝖺𝗇𝖽 𝗀𝖺𝗏𝖾 𝗍𝗁𝖾 𝖤𝖯𝖠 𝖺𝖼𝖼𝖾𝗌𝗌 𝗍𝗈 𝖠𝗎𝗍𝗈𝖽𝖾𝗌𝗄’𝗌 𝗈𝗇𝗅𝗂𝗇𝖾 𝖼𝗈𝗅𝗅𝖺𝖻𝗈𝗋𝖺𝗍𝗂𝗏𝖾 𝗉𝗅𝖺𝗍𝖿𝗈𝗋𝗆, 𝖠𝗎𝗍𝗈𝖽𝖾𝗌𝗄 𝖡𝖨𝖬 𝟥𝟨𝟢, for model sharing and project meetings, so they could rapidly model the 𝐚𝐬-𝐛𝐮𝐢𝐥𝐭 in their 𝐁𝐈𝐌 and 𝐂𝐀𝐃 solutions. 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐢𝐬 𝐌𝐚𝐧𝐝𝐚𝐭𝐨𝐫𝐲  They started to work with a company to create a 𝐁𝐈𝐌 𝐦𝐨𝐝𝐞𝐥 of the cathedral pre-fire based on laser scans created by art and architectural historian 𝐀𝐧𝐝𝐫𝐞𝐰 𝐓𝐚𝐥𝐥𝐨𝐧. As he had carried out a laser scan of Notre-Dame in 𝟤𝟢𝟣𝟢. With the backing of a European arts documentary, he set out to capture the cathedral from top to bottom using a Leica Geosystems laser scanner. 𝐏𝐫𝐞-𝐅𝐢𝐫𝐞 𝐁𝐈𝐌 𝐦𝐨𝐝𝐞𝐥 was created through a combination of laser scanning, photogrammetry, and drone footage, taking over a 𝐲𝐞𝐚𝐫 to complete. This effort resulted in a detailed 𝟑𝐃 𝐁𝐈𝐌 𝐦𝐨𝐝𝐞𝐥 of the structure as it appeared before the fire.  The EPA and 𝐀𝐮𝐭𝐨𝐝𝐞𝐬𝐤 collaborated on the development of the creation of a 𝐑𝐞𝐯𝐢𝐭 𝐁𝐈𝐌 𝐦𝐨𝐝𝐞𝐥, focusing on the structural elements of the building, the stonework. They modelled the crypt, the sacristy, the spire, Notre-Dame itself and the gardens behind the cathedral. There is no carpentry model because there were no detailed drawings of the roof structure and there were no laser scan in the voids. The post fire 𝐁𝐈𝐌 𝐦𝐨𝐝𝐞𝐥 was built in 𝐑𝐞𝐯𝐢𝐭 from a survey carried out using 𝟏𝟐 laser scanners which collected 𝟒𝟔,𝟎𝟎𝟎 scans. 𝐁𝐈𝐌 𝐁𝐞𝐧𝐞𝐟𝐢𝐭𝐬 The model was essential in addressing three major challenges: facilitating collaboration among more than 𝟐𝟎𝟎 𝐜𝐨𝐦𝐩𝐚𝐧𝐢𝐞𝐬 and 𝟐,𝟎𝟎𝟎 𝐢𝐧𝐝𝐢𝐯𝐢𝐝𝐮𝐚𝐥𝐬 involved in the project, meeting the ambitious five-year restoration timeline, and overcoming the limited documentation . 𝐓𝐡𝐞 𝐁𝐈𝐌 𝐦𝐨𝐝𝐞𝐥 also enabled precise planning, construction phase simulation, and early identification of potential issues, ensuring the project stayed on track. 𝐇𝐞𝐫𝐨𝐞𝐬 : Autodesk - Art Graphique & Patrimoine 𝐋𝐚𝐬𝐞𝐫 𝐒𝐜𝐚𝐧𝐧𝐢𝐧𝐠 𝟐𝟎𝟏𝟎 : Andrew Tallon -RIP- #bim #autodesk #revit #collaboration #architecture #engineering #construction #paris #france #cathedral

🛑 The Biggest Silent Killer of Mining Projects: Overconfidence in the Orebody 🛑 Every mine plan looks good... on paper. Production targets are met. Budgets approved. Equipment ordered. Everyone feels good until the mine starts underperforming. Month after month. Quarter after quarter. And the excuses pile up:   “Unexpected dilution” “Poor ground conditions” “Operational delays”   But here’s the truth nobody wants to say out loud:   The real failure happened years earlier, when we trusted the orebody model more than we should have. Mining is the only industry I know that builds billion-dollar businesses on statistical guesses... and then gets surprised when reality doesn't cooperate. Geological uncertainty is not a rounding error. It’s not a minor risk. It's shown to be the major contributor to project failures. It’s the foundation your entire operation stands on, or collapses on. And yet, companies build LOM plans assuming the estimated block model is the ground truth. Why? Because it's easier to assume certainty than to quantify uncertainty and plan for it. Because spreadsheets are cleaner when you don’t have multiple scenarios. Because no one wants to explain to the board that the “high-confidence” resource might still let them down. But pretending the orebody is perfect doesn't protect you. It just delays the realization.   🔍 Here’s what actually happens: Resource models, even “measured” ones, have built-in errors, including grade, volume, and continuity errors. Estimation methods like Kriging smooth out the grades, where high-grades (where we make money!) are underestimated, and low-grades are overestimated. Mine plans are optimized assuming every block behaves exactly as estimated. Operations find out the hard way that Mother Nature didn’t read the single 3D model.   🔴 And the cost? Missed production targets. Inability to control contaminants at the plant. Cash flow shortfalls. Poor reconciliation. Erosion of investor trust. Bad CAPEX decisions. Inability to fulfill contracts.   All because we decided to ignore the geological uncertainty!   ✅ What actually works? Quantify uncertainty, early and often. Simulate multiple orebody realizations that reproduce the local variability under the ground instead of relying on a single “best guess.” Optimize the strategic mine plan looking at all simulations. This will ensure you have integrated risk-management, prioritizing less risky, yet rich, areas early on till more information is available for later project stages. Report the production schedules probabilistically.   Mining doesn’t fail because it’s inefficient. It fails because it assumes the earth will behave the way a model says it should. And when that assumption breaks, everything else does too. Maybe it’s time we stop treating geological uncertainty as a technical inconvenience. It’s the core business risk, and facing it in advance is the only way we’ll stop falling short. #Uncertainty #Risk #ResourceModel #MinePlanning #Stochastic

Before construction, we build it in 3D. A building may look simple from the outside, but behind those walls there can be thousands of elements that all have to work together. Pipes. Ventilation. Electrical systems. Structural elements. Equipment. Fire protection. Access and maintenance zones. One small conflict between systems can become a serious problem once construction has already started. That's where BIM changes the process. Instead of discovering problems on site, teams can coordinate the building digitally first. MEP systems can be visualized, clashes detected, solutions tested, and different disciplines brought together in one environment. For me, one of the biggest advantages of BIM isn't simply creating a beautiful 3D model. It's the ability to think through the building before we build it. The earlier we find a problem, the easier and cheaper it is to solve. And this is where digital engineering becomes practical—not just a technology, but a way to reduce risk, improve coordination and make construction more predictable. Would you rather discover a design conflict on a computer screen—or on the construction site? #BIM #BIMModeling #Engineering #Architecture #MEP #Construction #DigitalEngineering #BuildingDesign #StructuralEngineering #ConstructionTechnology #PlanDi

Let's reverse engineer this demo. You need 3 things: (1) robust hardware and motor designs that treat simulation as first-class citizen; (2) a human motion capture ("mocap") dataset, such as those for film and gaming characters; (3) massively parallel RL training in GPU-accelerated simulation. Last October, our team trained a 1.5M parameter foundation model called HOVER for such agile motor control. It follows this recipe, roughly speaking: (1) Simulation used to be an after-thought. Now, it has to be part of the hardware design process. If your robot doesn't simulate well, you can kiss RL goodbye. Hardware-simulation co-design is a very interesting emergent topic that only becomes meaningful with today's compute capability. (2) Human mocap dataset to produce natural-looking walking and running gaits. That's one huge advantage of using humanoid robot - you get to imitate from tons of human motions that were originally captured for movies or AAA games. At least 3 ways to use the data: - For initialization: pre-train the neural net to imitate human, and then finetune it into the robot form factor with physics turned on; - For reward function: penalize any deviations from the target pose; - For representation learning: treat the human poses as a "motion prior" to constrain the space of robot behaviors. (3) Shove the above into Isaac sim, add a lot of randomization, pump it through PPO, throw in a bunch of GPUs, and then watch Netflix till loss converges. If you have an urge to comment this is CGI, let me save you a few keystrokes — many academic labs now own the G1 robot in the flesh. Read about our team's HOVER work: https://lnkd.in/gfKW9K5U

The teen models in Mango's latest campaign have perfect poses, perfect lighting, and one small detail: they don't exist. This Spanish fashion giant launched their Sunset Dream collection using entirely AI-generated models across 95 markets. Not a single human model was photographed. Here's how they did it: 📌 Took photos of real clothes on display stands 📌 Fed these pictures to their AI system 📌 Created model images in minutes 📌 Rolled out everywhere at once The business impact is massive. Fashion brands typically save 60-80% by leveraging AI photoshoots. Those savings can now fund innovation, better pricing, or faster expansion. But cost isn't the real story here. Speed is. While competitors wait weeks for campaign photos, MANGO creates, tests, and launches collections in days. No weather delays. No scheduling conflicts. No reshoots. This wasn't luck. Since 2018, Mango has built 15 different AI platforms across their business. They've been preparing for this moment. The result? Their 2024 turnover reached 3.3 billion euros in 2024, growing 7.6% from 2023. What makes this significant is that Mango proved AI-generated content can drive real sales. Their teen customers embraced these virtual models without hesitation. Fashion's biggest players are watching. If Mango's approach succeeds long-term, traditional photography could become a thing of the past for e-commerce. The brands that adapt now will set industry standards. Those that don't might find themselves competing against companies moving at AI speed. Which fashion tradition do you think AI will disrupt next?

Caltech has just flipped centuries of metalmaking on its head with a method that crafts alloys with surgical precision instead of brute force. By 3D-printing a hydrogel scaffold, soaking it in metallic salts, burning away the gel to leave metal oxides, and then stripping out oxygen in a hydrogen-rich environment, they can build parts atom by atom with exact compositions—no guesswork, no endless trial and error. The kicker? They can instantly fine-tune ratios—like shifting from a copper-heavy blend to a Cu12Ni88 mix nearly four times stronger—without altering the shape or tooling, turning metallurgy into a high-tech art form where strength, density, and performance are all set by design. https://lnkd.in/eVEZmsRZ FuturistSpeaker.com

Why Do We Use Fillets in High Stress Concentration Areas? In mechanical design, fillets play a crucial role in reducing stress concentration — especially at sharp internal corners where sudden geometry changes occur. High stress concentration areas are common failure points under load, making this design detail essential. In my FEA comparison study, I analyzed three cases: No Fillet – Extremely high stress near the sharp corner due to abrupt geometry transition. This is the most unsafe and prone to crack initiation. Small Fillet – Noticeable improvement in stress distribution, but still moderate localized stress. Large Fillet – Best performance with significantly smoother stress flow and minimal concentration. This results in increased durability and reliability. Why Large Fillets Are Safer: They distribute stress over a wider area. Minimize the peak stress values. Reduce the risk of fatigue failure and cracks. Improve component life under cyclic loading. Conclusion: When designing mechanical parts, especially those under dynamic or repeated loading, larger fillets at high-stress regions are a simple yet highly effective way to enhance strength and safety. . . . #FEA #StressAnalysis #MechanicalDesign #FiniteElementAnalysis #MechanicalEngineering #FilletDesign #DesignOptimization #EngineeringDesign #CAE #SolidMechanics #StressConcentration #FatigueAnalysis #ProductDesign #EngineeringTips #DesignForDurability #Simulation #StructuralAnalysis #CAD #EngineeringSimulation

Exploring different worlds in the browser with off-axis projection: Ancient tomb 🏺 → lava cavern 🌋 → overgrown ruins 🌿 Same structure, totally different feel. Here’s an experiment I've put together using Chisel from World Labs. Chisel lets you block out a 3D layout and generate a detailed world on top, so you define and block out the space first and then bring it to life. In this demo, I used that workflow to build and explore a range of environments with off-axis projection, all running in the browser with Three.js and Spark. Pipeline: → Block out structure in Blender or directly in Chisel → Use Chisel to generate a detailed environment from that layout + prompt → Bring it into Three.js and use Spark for rendering → Use Mediapipe for face tracking to enable off-axis projection You can go from a rough spatial idea to something you can actually move around in very quickly. Learn more about Chisel: https://lnkd.in/gJEewMcS