📢Diff3r: fast feed-forward 3DGS + per-scene optimization
Yueh-Cheng Liu predicts optimization-ready 3DGS init end to end, computing implicit gradients via Implicit Function Theorem + Gauss-Newton approximation for fast & stable results
Check it out: liu115.github.io/diff3r
📢Seen2Scene
Real-world 3D is incomplete, typically requiring training on synthetic scene data.
Quan Meng introduces visibility-guided flow matching, enabling training on real partial scans for scan completion & text-to-3D scene generation!
Check it out: quan-meng.github.io/projects/see...
📢Lookalike3D: Seeing Double in 3D
@cyeshwanth.bsky.social enables holistic, instance-consistent 3D object reconstruction & part segmentation by detecting identical and near-identical objects from multiview images.
Built on a dataset of 76k curated object pairs
cy94.github.io/lookalike3d/
Image & video synthesis struggle with the scale of truly large 3D scenes.
Manuel Schneider presents a geometry-first approach:
- geometric structure first: mesh scaffold defining the scene
- then appearance: mesh-conditioned image generation
Check it out: mschneider456.github.io/world-mesh/
The first keynote today at #CVMP2025 is by Angela Dai @adai.bsky.social from 3DAI Lab at TU Munich
"Can Transformers speak geometry?"
There are lots of different 3D representations we use in learning but industry widely uses meshes.
And four fantastic keynotes from Yi-Zhe Song @peoplecentredai.bsky.social, Christian Richardt @cr333.bsky.social , Angela Dai @adai.bsky.social , and Peter Hedman @peterhedman.bsky.social #CVMP2025
And that's a start 🤩
We kick off the 22nd ACM SIGGRAPH European Conference on Visual Media Production at BFI Southbank in London
Where industry and academia meet 🤝🎬
#CVMP2025 #BFI #BFISouthbank
📢ProcGen3D: Learning Neural Procedural Graphs for Image-to-3D Reconstruction
Xinyi Zhang learns neural procedural graphs to generate high-fidelity 3D - MCTS-guided sampling maintains consistency with the input image, even from real images!
Check it out: xzhang-t.github.io/project/Proc...
📢New in ScanNet++: High-Res 360° Panos!
Chandan Yeshwanth and Yueh-Cheng Liu have added pano captures for 956 ScanNet++ scenes, fully aligned with the 3D meshes, DSLR, and iPhone data - multiple panos per scene
Check it out:
Docs kaldir.vc.in.tum.de/scannetpp/do...
Code github.com/scannetpp/sc...
📢 Excited to share our latest #ICCV2025 work DiffuMatch: learning spectral diffusion priors for robust non-rigid shape matching! (1/n)
for more documentation: github.com/scannetpp/sc...
Huge thanks to Yueh-Cheng Liu, as well as Chandan Yeshwanth and @niessner.bsky.social for their incredible work!
📢 📢 📢 We've released the ScanNet++ Novel View Synthesis Benchmark for iPhone data!
Test your models on RGBD video featuring real-world challenges like exposure changes & motion blur!
Download the newest iPhone NVS test split and submit your results! ⬇️
scannetpp.mlsg.cit.tum.de/scannetpp/be...
🚀We have PhD openings in my lab at TU Munich!
Explore 3D/4D reconstruction & generation, semantic & functional understanding, and more - at the intersection of graphics, vision, and machine learning.
💼PhDs are 100% E13 positions
👉Apply: 3dunderstanding.org/openings.html or via ELLIS!
Excited to join @uvadatascience.bsky.social as an Assistant Professor!
Deeply grateful to my advisors @adai.bsky.social, Maks Ovsjanikov, Hongbo Fu, and Chiew-Lan Tai for their unwavering support.
📣We are recruiting PhD students and postdocs to work on #SpatialAI. Flyer below with details!
Super excited to have @phillipisola.bsky.social give a guest lecture in our Introduction to Deep Learning course at TU Munich on representations learned by neural networks!
Check out our #ICCV2025 work on functional 3d scan editing, learning to optimize, multi-level 3d captioning, interactive mesh editing, audio-driven avatars, & shape matching!
Congrats to Amine, Yueh-Cheng, @cyeshwanth.bsky.social, Haoxuan, Shivangi, and Emery for their amazing work!
Check out Xinyi’s super cool work on 4D generative modeling with dictionary neural fields - also at poster 12 in poster session 6 today! #cvpr2025
Fascinating lessons on 3D reconstruction from Andrea Vedaldi @oxford-vgg.bsky.social at the ScanNet++ workshop @cvprconference.bsky.social!
Learning about world models with Gordon Wetzstein at the ScanNet++ workshop @cvprconference.bsky.social!
Super cool insights on 4D reconstruction from Qianqian Wang at the ScanNet++ workshop @cvprconference.bsky.social!
How to generate 3D with LLM reasoning by Cordelia Schmid at the ScanNet++ workshop @cvprconference.bsky.social!
Kicking off the ScanNet++ workshop @cvprconference.bsky.social with Katja Schwarz on how to generate 3D with diffusion models!
Check out the ScanNet++ workshop @CVPR on June 12 in 211 from 8:50am!
Exciting keynotes on state-of-the-art NVS & 3D understanding from Andrea Vedaldi, Cordelia Schmid, Gordon Wetzstein, Katja Schwarz, Qianqian Wang, and leading methods on the benchmark!
kaldir.vc.in.tum.de/scannetpp/cv...
🚀🚀🚀Announcing our $13M funding round to build the next generation of AI: 𝐒𝐩𝐚𝐭𝐢𝐚𝐥 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥𝐬 that can generate entire 3D environments anchored in space & time. 🚀🚀🚀
Interested? Join our world-class team:
🌍 spaitial.ai
youtu.be/FiGX82RUz8U
Brilliant insights from @michael-j-black.bsky.social on the importance of data and 3D+ for 4D foundation models that understand humans, and the future of embodied intelligence in the last keynote talk of #Eurographics2025!
See you next year in Aachen :)
Awesome look into the future of humanoid robots and what we can learn from character animation from Karen Liu’s keynote at #Eurographics2025!
Amazing keynote by Alyosha Efros on the role of data in visual computing at #Eurographics2025!
Thought-provoking insights from generative models to 3D perception :)
What an amazing start to our #Eurographics2025 keynotes from @justinmsolomon.bsky.social on optimization for geometry processing - lots of useful optimization tools, and so many more interesting challenges yet to be solved!
📢𝐐𝐮𝐢𝐜𝐤𝐒𝐩𝐥𝐚𝐭: Fast 3D Surface Reconstruction via Learned Gaussian Initialization
Yueh-Cheng Liu learns 2DGS initialization, densification, and optimization priors from ScanNet++ => fast & accurate reconstruction!
Project: liu115.github.io/quicksplat
w/ Lukas Höllein and @niessner.bsky.social
