Chin-Yang Lin

Ph.D. student @ NYCU | Research Intern @ NVIDIA

[Chin-Yang Lin @linjohnss]

🔥 Robust 3D Gaussian Splatting 🔥

#LongSplat turns any casual long video into a coherent 3D world 🌍
It reconstructs scenes without camera poses and renders high-quality novel views 🎥

Project: https://t.co/fTWxPfEfw2
Code: https://t.co/FjcewUb9fe

[NVIDIA AI Developer @NVIDIAAIDev]

Take a video of your favorite trail and turn it into a coherent 3D world with LongSplat 🤳⛰️

Our model reconstructs scenes from any casual long video without camera calibration and renders high-quality novel views from any point along your path.

Try it for yourself ➡️ https://t.co/0yD1LYL8IS

[DailyPapers @HuggingPapers]

Scal3R from NVIDIA

Accurate online 3D reconstruction on kilometer-scale sequences with only ~1% extra parameters on a frozen backbone, trained in 8 hours on a single GPU.

[Zhenjun Zhao @zhenjun_zhao]

Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction

@linjohnss, Yang-Che Sun, Cheng Sun, @FuEnYang1, @CMHungSteven, Yen-Yu Lin, Wei-Chen Chiu, @YuLunLiu790407

tl;dr: relative pose tokens->queries; image tokens->self-attention

https://t.co/d90mi6l0Fc

[European Conference on Computer Vision #ECCV2026 @eccvconf]

Thank you to all our #ECCV2026 reviewers. Special thanks to the Outstanding Reviewers for their exceptional contributions! 👏

https://t.co/LUZEHdAZ6z

[Zhenjun Zhao @zhenjun_zhao]

AdaGaR: Adaptive Gabor Representation for Dynamic Scene Reconstruction

@JiewenChan, @zhenjun_zhao, @YuLunLiu790407

tl;dr: extend Gaussians to frequency domain->hybrid Gabor & Gaussian->novel video representation->adaptive high & low-frequency balance

https://t.co/CBhv3fzWZw

[AK @_akhaliq]

Stream-DiffVSR

Low-Latency Streamable Video Super-Resolution via Auto-Regressive Diffusion

[Wildminder @wildmindai]

Stream-DiffVSR: A low-latency, streamable video super-resolution model;
- distills Stable Diffusion Upscaler to just 4 steps;
- processes 720p ~3fps.
https://t.co/67a63cn1QJ

[Chin-Yang Lin @linjohnss]

🎥 Some results from LongSplat!
After reconstructing the 3D scene and camera poses, you can render any viewpoint along the path, even achieve video stabilization.

[Chin-Yang Lin @linjohnss]

🚀Excited to share our #ICCV2025 paper with NVIDIA Research
LongSplat: Robust Unposed 3D Gaussian Splatting for Casual Long Videos
Enables stable, high-quality 3D reconstruction even from long or shaky handheld videos!
New YouTube video drop👇
https://t.co/If09qZ8Sk4

[Chin-Yang Lin @linjohnss]

📄Paper：https://t.co/v2h3Y1kuB9
🌐Project Page：https://t.co/yqIaPmtEKt
🎬Video：https://t.co/If09qZ8Sk4
⌨️Code：https://t.co/SA46CjyCz6 (⭐️Star it!)

🌺Meet us at #ICCV2025:
Oct 20 10:20–11:00 – Wild-3D Workshop (Ex Hall II)
Oct 23 14:30–16:30 – Main Poster #259 (Ex Hall I)

[MrNeRF @janusch_patas]

Code dropped:
https://t.co/fiIPX7TGH3

[Donghoon Ahn @donghoon_ahn]

By randomly walking the campus, I encountered something surprising, the real NeRF scene! It's awesome!!

[AK @_akhaliq]

LongSplat

Robust Unposed 3D Gaussian Splatting for Casual Long Videos

[DailyPapers @HuggingPapers]

NVIDIA presents LongSplat:
Robust Unposed 3D Gaussian Splatting for Casual Long Videos!

Reconstruct expansive scenes from any video, tackling pose drift & memory limitations for SOTA results.

[Chin-Yang Lin @linjohnss]

Thanks for the shout-out @zhenjun_zhao! 🙌 Really appreciate you sharing our work.
Check out the project page: https://t.co/lrMAml5Uf0
Code: https://t.co/SA46CjyCz6

[Zhenjun Zhao @zhenjun_zhao]

LongSplat: Robust Unposed 3D Gaussian Splatting for Casual Long Videos

@linjohnss, Cheng Sun, @FuEnYang1, @CMHungSteven, Yen-Yu Lin, @YuLunLiu790407

tl;dr: MASt3R+octree from F2-NeRF+Scaffold-GS for initialization; MASt3R+PnP+RANSAC for pose estimation; visibility local opt.+global opt.

https://t.co/hslZ0wl3as

[Chin-Yang Lin @linjohnss]

@yu47257737 @janusch_patas Apologies for the oversight, and thank you for pointing it out. I have now included the proper website citation to acknowledge your work🙏.

[Chin-Yang Lin @linjohnss]

Thank you @janusch_patas for highlighting our work LongSplat!

[MrNeRF @janusch_patas]

LongSplat: Robust Unposed 3D Gaussian Splatting for Casual Long Videos

Contributions:
• An incremental joint optimization approach for simultaneous camera pose and 3DGS reconstruction, reducing local minima and ensuring global consistency.

• A robust pose estimation module leveraging learned 3D priors for accurate camera pose estimation.

• An adaptive Octree Anchor Formation strategy that significantly reduces memory usage while preserving reconstruction quality.

[MrNeRF @janusch_patas]

LongSplat: Robust Unposed 3D Gaussian Splatting for Casual Long Videos

Contributions:
• An incremental joint optimization approach for simultaneous camera pose and 3DGS reconstruction, reducing local minima and ensuring global consistency.

• A robust pose estimation module leveraging learned 3D priors for accurate camera pose estimation.

• An adaptive Octree Anchor Formation strategy that significantly reduces memory usage while preserving reconstruction quality.

[Chin-Yang Lin @linjohnss]

🚀 Excited to share that our paper 𝗙𝗿𝘂𝗴𝗮𝗹𝗡𝗲𝗥𝗙: 𝗙𝗮𝘀𝘁 𝗖𝗼𝗻𝘃𝗲𝗿𝗴𝗲𝗻𝗰𝗲 𝗳𝗼𝗿 𝗘𝘅𝘁𝗿𝗲𝗺𝗲 𝗙𝗲𝘄-𝘀𝗵𝗼𝘁 𝗡𝗼𝘃𝗲𝗹 𝗩𝗶𝗲𝘄 𝗦𝘆𝗻𝘁𝗵𝗲𝘀𝗶𝘀 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗟𝗲𝗮𝗿𝗻𝗲𝗱 𝗣𝗿𝗶𝗼𝗿𝘀 has been accepted to the #CVPR2025 Main Conference! 🎉

[Zhenjun Zhao @zhenjun_zhao]

AuraFusion360: Augmented Unseen Region Alignment for Reference-based 360° Unbounded Scene Inpainting

@kwlilspace, Yang-Jung Chen, Ying-Huan Chen, @jayinnn, Bo-Hsu Ke, Chun-Wei Tuan Mu, Yi-Chuan Huang, Chin-Yang Lin, @CMHungSteven, Yen-Yu Lin, Yu-Lun Liu