Bernhard Kerbl
@Snosixtytwo
Followers
1K
Following
80
Media
15
Statuses
164
Visiting Researcher at Carnegie Mellon University
Pittsburgh, Pennsylvania
Joined September 2009
This was a great effort by the young team at @CarnegieMellon , supervised by @FerranDeLaTorre , and Francisco Carrasco Vicente, led by @saswat_mallick , Junkai Huang, with Marc Ruiz Olle, @claansy_ , and Albert Mosella-Montoro.
0
0
1
Our approach to interactive reconstruction uses distributed GPUs & coarse-to-fine optimization, with photorealistic results that scale effortlessly. The BASKET-Multiview tests fidelity in true 3D, making it an exceptional benchmark for evaluating 4D scene reconstruction quality.
1
0
1
We introduce BASKET-Multiview. A synthetic, multi-view dynamic dataset capturing complex human motions in sports events with 360° views and full ground-truth annotations (RGB, depth, normals, masks, and animation). Find it on the project site!
1
0
1
What's in it?.1. High-fidelity, interactive 6-DOF viewing .2. Distributed & scalable 4D reconstruction .3. Outperforms all SOTA methods in speed and quality.4. Comes with BASKET-Multiview, the most detailed multi-view sports dataset yet.
1
0
1
SIGGRAPH is over, but the papers remain! .If you missed it, our team presented Echoes of the Coliseum: Towards 3D Live Streaming of Sports Events. Check it out here:.
4
13
79
Janusch, take the wheel.
🚀 Want to contribute to the fastest craziest 3D Gaussian Splatting repo out there?. We are moving fast to beat every commercial splatting software and it will stay free forever under GPLv3. I listed some GitHub issues to help you get started. Get your toes wet! Be part of it!.
0
0
7
RT @AndreasMeuleman: We are very excited to announce our work "On-the-fly Reconstruction for Large-Scale Novel View Synthesis from Unposed….
0
66
0
RT @janusch_patas: StochasticSplats: Stochastic Rasterization for Sorting-Free 3D Gaussian Splatting. Abstract:.3D Gaussian splatting (3DGS….
0
13
0
Two days after the h-1b fallout, and 80% of the nazi stuff has vanished from my twitter feed. Interesting?.
1
0
3
RT @PeterHedman3: This project blew my mind! It looks just as good as 3DGS but it has no per-scene optimization or reconstruction. Every fr….
0
17
0
The original 3DGS code is an academic research project, the fact that somebody is still actively improving it 1.5 years later is 🧑🍳😘 (chef's kiss).
Some new updates were quietly pushed on the inria-3dgs repository:.1. Integration of the Taming3DGS.2. Integration of pre-filtering similar to Mip-Splatting.3. Depth regularization and Exposure regularization.4. QoL improvements on the sIBR viewer. Nice work from @alanvinx.
0
0
45
Hot dang!.
Hate waiting 10 minutes for 3D GS to render your favorite indoor or outdoor scenes? ⏳ Our feed-forward solution, Long-LRM, cuts it down to just 1 second! ⚡️ With a straightforward mix of Mamba2 and transformer, it scales up to 32 high-res input images.
0
0
6
RT @fhahlbohm: Here is our shot at reducing artifacts in 3D Gaussian splat rendering. Paper: Website: https://t.co….
0
22
0
RT @_linus_franke: Our latest work to increase multi-view coherence in 3D Gaussian Splatting using Hybrid Transparency:. - Perspective-corr….
0
10
0
Awesome work just dropped for high-quality relightable Gaussians. Huge congrats to @GinterYohan !.
The code for our project *A Diffusion Approach to Radiance Field Relighting using Multi-Illumination Synthesis* is now available at: Project Page: Thanks to the amazing work from @GinterYohan you can now relight your 3DGS!
0
0
1
I myself am currently very much looking for faculty positions! It's been a productive year so far - here's my homepage, if you think I might be a good fit, my comms are very open:.👉
0
1
1
As part of this work, Rahul made a lightning-fast SSIM implementation, which you can find here: Rahul is now looking for PhD positions! If you need someone with an outstanding record and the skills + initiative to do amazing work.👉
github.com
Lightning fast differentiable SSIM. Contribute to rahul-goel/fused-ssim development by creating an account on GitHub.
2
0
10
This was a really efficient, synergistic collaboration between the folks from @tugraz and @CarnegieMellon thanks to the guidance of Markus Steinberger, Francisco Carrasco Vicente, and @FerranDeLaTorre.
2
0
0
Code for our optimizations is available now as a delta to the INRIA codebase. More coming soon. Expect to see it integrated into official 3DGS repositories as well!.
humansensinglab.github.io
2
1
5
Not only can you get high-quality reconstructions that are 5x smaller, but also get them 5x faster! This is due to careful performance optimizations: we analyzed the main bottlenecks and redesigned the slowest steps, like the backward logic.
1
0
1