iiinova — where your digital self keeps evolving

Haofei Xu @haofeixu.bsky.social · 27/03/2026

It's incredibly exciting to see these old ideas shine again, empowered by modern architectures and tools. It’s equally rewarding to watch young students grow, and I'm so happy to be able to support their research journey!

000

Haofei Xu @haofeixu.bsky.social · 27/03/2026

A little backstory: we developed GMFlow (CVPR 2022, github.com/haofeixu/gmflow) and UniMatch (TPAMI 2023, github.com/autonomousvisi…) about 4 years ago.

100

Haofei Xu @haofeixu.bsky.social · 27/03/2026

This project was done by our truly exceptional Master's student @dingxi-z.bsky.social (an incoming PhD student at ETH and MPI), together with amazing @fangjinhuawang.bsky.social and @marcpollefeys.bsky.social !

100

Haofei Xu @haofeixu.bsky.social · 27/03/2026

🏆 SOTA zero-shot performance across multiple benchmarks and in-the-wild videos. Check out the links below to see it in action! 👇 🌐 Project: kristen-z.github.io/projects/meg... 📄 Paper: arxiv.org/abs/2603.25739 💻 Code: github.com/cvg/megaflow

100

Haofei Xu @haofeixu.bsky.social · 27/03/2026

🎯 Large displacement motion estimation (optical flow, point tracking) has been a long-standing challenge. MegaFlow proposes a simple solution by leveraging pre-trained vision priors. Combined with a global matching formulation, we tackle both optical flow and point tracking in a unified structure.

100

Haofei Xu @haofeixu.bsky.social · 27/03/2026

🚀 Introducing MegaFlow: Zero-Shot Large Displacement Optical Flow! 🔥 Code, models, and demo — all available now!

110

Haofei Xu @haofeixu.bsky.social · 10/10/2025

🏆ReSplat achieves state-of-the-art performance on DL3DV and RealEstate10K, across various input views (2, 8, 16) and resolutions (256x256-540×960). Check out our paper for detailed results: arxiv.org/abs/2510.08575

110

Haofei Xu @haofeixu.bsky.social · 10/10/2025

🧠To initialize the recurrent process, we design a compact reconstruction model that operates in a 16x subsampled space, producing 16x fewer Gaussians than previous pixel-aligned models. This substantially reduces computational overhead and allows for efficient Gaussian updates.

110

Haofei Xu @haofeixu.bsky.social · 10/10/2025

🎯Key idea: the Gaussian splatting rendering error provides a rich feedback signal, guiding the recurrent network to learn effective Gaussian updates. This feedback naturally adapts to unseen data, enabling robust generalization across datasets and resolutions.

110

Haofei Xu @haofeixu.bsky.social · 10/10/2025

⚡Feed-forward Gaussian splatting is fast but limited: it only makes a single forward pass. ReSplat introduces recurrent refinement, enabling the model to iteratively improve the 3D Gaussians. ReSplat converges fast, requiring only three iterations.

110

Haofei Xu @haofeixu.bsky.social · 10/10/2025

🚀Excited to share our recent work on test-time scaling for feed-forward Gaussian splatting: we learn a recurrent model ReSplat that is able to iteratively improve the reconstruction quality in a feed-forward manner! haofeixu.github.io/resplat/

1103

Haofei Xu @haofeixu.bsky.social · 29/08/2025

Check out Frano's amazing work on multi-view 3D point tracking! 🚀 Code, models, datasets, and interactive results — all available!

020

Reposted by Haofei Xu

haoyuhe.bsky.social @haoyuhe.bsky.social · 20/08/2025

🚀 Introducing our new paper, MDPO: Overcoming the Training-Inference Divide of Masked Diffusion Language Models. 📄 Paper: www.scholar-inbox.com/papers/He202... arxiv.org/pdf/2508.13148 💻 Code: github.com/autonomousvi... 🌐 Project Page: cli212.github.io/MDPO/

1128

Haofei Xu @haofeixu.bsky.social · 14/06/2025

Project page: chengzhag.github.io/publication/... Code: github.com/chengzhag/Pa... Interact with the scene in this video (best viewed on a desktop browser or Youtube app): www.youtube.com/watch?v=9bKZ...

010

Haofei Xu @haofeixu.bsky.social · 14/06/2025

Wanna scale your feed-forward Gaussian Splatting model to 4K resolution? Come check out our #CVPR2025 poster PanSplat today 10:30–12:30 (June 14) at ExHall D, Poster #74!

140

Haofei Xu @haofeixu.bsky.social · 05/06/2025

Catch us on Saturday, June 14 at 5 PM, ExHall D Poster #58!

000

Reposted by Haofei Xu

Andreas Geiger @andreasgeiger.bsky.social · 05/06/2025

Your personalized CVPR 25 @cvprconference.bsky.social conference programs are now available for you! www.scholar-inbox.com/conference/c...

15416

Haofei Xu @haofeixu.bsky.social · 05/06/2025

DepthSplat: Connecting Gaussian Splatting and Depth Project page: haofeixu.github.io/depthsplat/ Code, models, data: github.com/cvg/depthsplat

000

Haofei Xu @haofeixu.bsky.social · 05/06/2025

Excited to present our #CVPR2025 paper DepthSplat next week! DepthSplat is a feed-forward model that achieves high-quality Gaussian reconstruction and view synthesis in just 0.6 seconds. Looking forward to great conversations at the conference!

3267

Reposted by Haofei Xu

Katrin Renz @katrinrenz.bsky.social · 08/05/2025

📣 Excited to share our #CVPR2025 Spotlight paper and my internship project @wayve: SimLingo. A Vision-Language-Action (VLA) model that achieves state-of-the-art driving performance with language capabilities. Code: github.com/RenzKa/simli... Paper: arxiv.org/abs/2503.09594

1259

Reposted by Haofei Xu

Stefano Esposito @s-esposito.bsky.social · 05/05/2025

📢 New paper CVPR 25! Can meshes capture fuzzy geometry? Volumetric Surfaces uses adaptive textured shells to model hair, fur without the splatting / volume overhead. It’s fast, looks great, and runs in real time even on budget phones. 🔗 autonomousvision.github.io/volsurfs/ 📄 arxiv.org/pdf/2409.02482

13221

Reposted by Haofei Xu

Bernhard Jaeger @bernhard-jaeger.bsky.social · 28/04/2025

Introducing CaRL: Learning Scalable Planning Policies with Simple Rewards We show how simple rewards enable scaling up PPO for planning. CaRL outperforms all prior learning-based approaches on nuPlan Val14 and CARLA longest6 v2, using less inference compute. arxiv.org/abs/2504.17838

02414

Reposted by Haofei Xu

Andreas Geiger @andreasgeiger.bsky.social · 24/04/2025

🏠 Introducing DepthSplat: a framework that connects Gaussian splatting with single- and multi-view depth estimation. This enables robust depth modeling and high-quality view synthesis with state-of-the-art results on ScanNet, RealEstate10K, and DL3DV. 🔗 haofeixu.github.io/depthsplat/

13913

Reposted by Haofei Xu

Haiwen Huang @haiwen-huang.bsky.social · 22/04/2025

Excited to introduce LoftUp! A strong (than ever) and lightweight feature upsampler for vision encoders that can boost performance on dense prediction tasks by 20%–100%! Easy to plug into models like DINOv2, CLIP, SigLIP — simple design, big gains. Try it out! github.com/andrehuang/l...

0195

Reposted by Haofei Xu

Kashyap Chitta @kashyap7x.bsky.social · 17/03/2025

🐎 Centaur, our first foray into test-time training for end-to-end driving. No retraining needed, just plug-and-play at deployment given a trained model. Also, theoretically nearly no overhead in latency with some clever use of buffers. Surprising how effective this is! arxiv.org/abs/2503.11650

1127

Reposted by Haofei Xu

Andreas Geiger @andreasgeiger.bsky.social · 16/01/2025

This week we had our winter retreat jointly with Daniel Cremer's group in Montafon, Austria. 46 talks, 100 Km of slopes and night sledding with some occasionally lost and found. It has been fun!

07211