[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

[@janusch_patas](/creator/twitter/janusch_patas)
"🔔 Bounty #004 - Automatic Per-Scene Hyperparameter Optimization Build a system for gplv3 LichtFeld Studio that automatically finds optimal hyperparams per scene during training. Must use MCMC densification (Num Gaussians tunable). Target: +0.15 dB PSNR over MipNeRF360 baseline. Prize pool: $2430. Details and link in the thread below"  
[X Link](https://x.com/janusch_patas/status/1967278789504983412) [@janusch_patas](/creator/x/janusch_patas) 2025-09-14T17:26Z 14.4K followers, 28.5K engagements


"Fix False Transparency by Noise Guided Splatting Contributions: X. A new technique that places interior noise Gaussians to distinguish between the interior and exterior space of the rendered object guiding optimization toward proper surface opacity. This method is plug-and-play and requires minimal modifications to existing frameworks. X. A new approach to visualize and quantify false transparency in static 3DGS renderings. X. A noise Gaussian infill dataset including infill add-ons to existing datasets (e.g. DTU) and a customized high-resolution object-centric scan dataset to facilitate"  
[X Link](https://x.com/janusch_patas/status/1980165494767501515) [@janusch_patas](/creator/x/janusch_patas) 2025-10-20T06:53Z 14.4K followers, 3460 engagements


"I love the look when frustums are added and the initial point cloud is visualized. The points are shaded voxels. Coming soon (today) in LichtField Studio"  
[X Link](https://x.com/janusch_patas/status/1962815475798417417) [@janusch_patas](/creator/x/janusch_patas) 2025-09-02T09:50Z 14.4K followers, 4625 engagements


"@YeheLiu You are welcome. Note some links are broken on the project page"  
[X Link](https://x.com/janusch_patas/status/1980226588571799702) [@janusch_patas](/creator/x/janusch_patas) 2025-10-20T10:56Z 14.4K followers, XX engagements


"GeoSVR: Taming Sparse Voxels for Geometrically Accurate Surface Reconstruction Contributions: An exploration of GeoSVR to build an explicit voxel-based framework for accurate surface reconstruction taming sparse voxels to enable delicate and complete geometry learning. A Voxel-Uncertainty Depth Constraint that maximizes the utilization of external depth cues while avoiding quality degradation through the proposed voxel uncertainty evaluation enabling effective and robust scene constraints for highly accurate surface reconstruction. A Sparse Voxel Surface Regularization for surface geometric"  
[X Link](https://x.com/janusch_patas/status/1970517063908413512) [@janusch_patas](/creator/x/janusch_patas) 2025-09-23T15:54Z 14.4K followers, 3525 engagements


"MaterialRefGS: Reflective Gaussian Splatting with Multi-view Consistent Material Inference Contributions: We propose a novel approach to modeling reflections through Gaussian Splatting with multi-view consistent material inference. This includes a multi-view material consistency constraint and reflection strength prior supervision. Our approach offers a new perspective for modeling reflections via physically grounded illumination decomposition. We introduce a differentiable environment modeling strategy using 2DGS-based ray tracing enhancing photorealistic rendering of indirectly illuminated"  
[X Link](https://x.com/janusch_patas/status/1978352134413562095) [@janusch_patas](/creator/x/janusch_patas) 2025-10-15T06:47Z 14.4K followers, 3642 engagements


"InstantSfM: Fully Sparse and Parallel Structure-from-Motion TLDR: InstantSfM is a fully sparse and parallel Structure-from-Motion pipeline. It leverages GPU acceleration to achieve up to XX speedup over traditional methods like COLMAP while maintaining or improving reconstruction accuracy across diverse datasets. Contributions: We extend sparse-aware bundle adjustment techniques to global positioning introducing a complete global SfM system in PyTorch. We demonstrate state-of-the-art efficiency while achieving comparable accuracy to both traditional well-established SfM pipelines and"  
[X Link](https://x.com/janusch_patas/status/1978715515888832564) [@janusch_patas](/creator/x/janusch_patas) 2025-10-16T06:51Z 14.4K followers, 23.3K engagements


"3D Gaussian Splatting Driving Simulator You know those YouTube walking tours I think this could become a similar trend like driving around in a splat or taking your driver's license test in a 3DGS simulator"  
[X Link](https://x.com/janusch_patas/status/1979215379005345959) [@janusch_patas](/creator/x/janusch_patas) 2025-10-17T15:58Z 14.4K followers, 136.9K engagements


"Thats crazy. 1.5km 3D Gaussian Driving at 280-500fps. City scale Gaussians with exciting real world applications"  
[X Link](https://x.com/janusch_patas/status/1979540209655316646) [@janusch_patas](/creator/x/janusch_patas) 2025-10-18T13:28Z 14.4K followers, 47.3K engagements


"Pretty cool DIY setup for scanning roads for 3D Gaussian Splatting This dude is crushing it"  
[X Link](https://x.com/janusch_patas/status/1979824800475541755) [@janusch_patas](/creator/x/janusch_patas) 2025-10-19T08:19Z 14.4K followers, 6160 engagements


"Skyfall-GS: Synthesizing Immersive 3D Urban Scenes from Satellite Imagery TL;DR: Skyfall-GS converts satellite images to explorable 3D urban scenes using diffusion models with real-time rendering performance. Contributions: We introduce Skyfall-GS the first method to synthesize immersive real-time free-flight navigable 3D urban scenes solely from multi-view satellite imagery using generative refinement. An open-domain refinement approach leverages pre-trained text-to-image diffusion models without domain-specific training. A curriculum-learning-based iterative refinement strategy"  
[X Link](https://x.com/janusch_patas/status/1980163372160344212) [@janusch_patas](/creator/x/janusch_patas) 2025-10-20T06:45Z 14.4K followers, 26.3K engagements


"GaussGym: An open-source real-to-sim framework for learning locomotion from pixels Contributions: X. GaussGym: A fast open-source photorealistic simulator with 2500 scenes. It supports diverse scene creation from manual scans open-source datasets and generative video models. X. We share findings on addressing the visual sim-to-real gap showing that incorporating geometry reconstruction as an auxiliary task significantly improves stair-climbing performance. X. We demonstrate the semantic reasoning of RGB navigation policies in a goal-reaching task where policies trained on pixels successfully"  
[X Link](https://x.com/janusch_patas/status/1980167617362141470) [@janusch_patas](/creator/x/janusch_patas) 2025-10-20T07:02Z 14.4K followers, 8760 engagements


"3D bee scan magic: captured in stunning detail with Gaussian splatting using just XXX images"  
[X Link](https://x.com/janusch_patas/status/1980638032526225607) [@janusch_patas](/creator/x/janusch_patas) 2025-10-21T14:11Z 14.4K followers, 32.7K engagements