𝗕𝗹𝗼𝗰𝗸𝟯𝗗: 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗧𝗲𝘅𝘁-𝘁𝗼-𝟯𝗗 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 𝘃𝗶𝗮 𝗕𝗹𝗼𝗰𝗸-𝗪𝗶𝘀𝗲 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻
Bowen Cui, Weijie Wang, Zeyu Zhang ... Bohan Zhuang
arxiv.org/abs/2608.19567
Trending on scholar-inbox.com

🚀 Introducing World-R1: Video models already know 3D — they just need RL to wake it up!
No arch changes. No video training data. No extra inference cost.⬇️
🌐Website: aka.ms/world-r1

📢 We release "Feed-Forward 3D Scene Modeling: A Problem-Driven Perspective" — a comprehensive survey covering 200+ papers on feed-forward 3D reconstruction!
Instead of categorizing by 3D representations, we propose a problem-driven taxonomy.
🌐 ff3d-survey.github.io

Thanks @ducha-aiki.bsky.social for sharing our work, we truly appreciate it! More related links are listed below, and we welcome the community's feedback.

Project Page: depth-anything-3.github.io
Code: github.com/ByteDance-Se...
Hugging Face Demo: huggingface.co/spaces/depth...

Trace Anything: Representing Any Video in 4D via Trajectory Fields

Xinhang Liu, Yuxi Xiao, @donydchen.bsky.social, Jiashi Feng, Yu-Wing Tai, Chi-Keung Tang, Bingyi Kang

arxiv.org/abs/2510.13802

𝗩𝗼𝗹𝗦𝗽𝗹𝗮𝘁: 𝗥𝗲𝘁𝗵𝗶𝗻𝗸𝗶𝗻𝗴 𝗙𝗲𝗲𝗱-𝗙𝗼𝗿𝘄𝗮𝗿𝗱 𝟯𝗗 𝗚𝗮𝘂𝘀𝘀𝗶𝗮𝗻 𝗦𝗽𝗹𝗮𝘁𝘁𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗩𝗼𝘅𝗲𝗹-𝗔𝗹𝗶𝗴𝗻𝗲𝗱 𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗼𝗻
Weijie Wang, Yeqing Chen, Zeyu Zhang ... Bohan Zhuang
arxiv.org/abs/2509.19297
Trending on www.scholar-inbox.com

𝗥𝗲𝘃𝗶𝘀𝗶𝘁𝗶𝗻𝗴 𝗗𝗲𝗽𝘁𝗵 𝗥𝗲𝗽𝗿𝗲𝘀𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻𝘀 𝗳𝗼𝗿 𝗙𝗲𝗲𝗱-𝗙𝗼𝗿𝘄𝗮𝗿𝗱 𝟯𝗗 𝗚𝗮𝘂𝘀𝘀𝗶𝗮𝗻 𝗦𝗽𝗹𝗮𝘁𝘁𝗶𝗻𝗴
Duochao Shi, Weijie Wang, Donny Y. Chen ... Chunhua Shen
arxiv.org/abs/2506.05327
Trending on www.scholar-inbox.com

Thanks Zhenjun for sharing. Our PM-Loss tackles low geometry quality in feed-forward 3DGS caused by "discontinuity" in predicted depth, and features a plug-and-play training loss that leverages geometry priors from pointmap-based models like VGGT. More at aim-uofa.github.io/PMLoss

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!

Thanks Zhenjun for sharing. Our work presents ZPressor, an architecture-agnostic module that enables existing feed-forward #3DGS models to effectively handle dense input views (up to 100 on an 80G GPU), drawing inspiration from the Information Bottleneck principle. More at bsky.app/profile/weij....

Very happy to share donydchen.github.io/mvsplat360/ at #3DV2025 Nectar Track.