Vincent Qin

⭐️Focusing on Visual Localization, SfM and SLAM.

[Vincent Qin @AlphaRealcat]

Image matching webui is now deployed on HF, link: https://t.co/WhpfeO7wcc

[Nicky Kriplani @NickyDCFP]

Joint work with @YiwenZhangYZ, @ang3linawang, and @ElorHadar at Cornell Tech.

Paper: https://t.co/jDrHfiIUKX
Project Page: https://t.co/59aOWjpAhN
Code: https://t.co/ikvQrUFvnO
Data: https://t.co/UhiKg9P98H

[Vincent Qin @AlphaRealcat]

Code: https://t.co/nNQZTCuPF0
Project: https://t.co/I3pg9PeQGg

[Zhenjun Zhao @zhenjun_zhao]

TartanMatch: Towards Universal Dense Correspondence Across Modalities

Hyeokjoon Kwon, Jiting Cai, Ruogu Li, Kritan Bhandari, Geethika Hemkumar, Parv Maheshwari, @QiuYuhengQiu, @YuchenZhan54250, @smash0190, Wenshan Wang

tl;dr: multi-modal UFM

https://t.co/Bgtjh8FJPH

[Nick @nickcammarata]

interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had

with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works

https://t.co/zHij5zwKGT

[Zhenjun Zhao @zhenjun_zhao]

MIRAGE: Metric Inference and Reliability-Aware Gaussian Estimation from Feed-Forward Vision

Xinzhe Wang, Changjian Jiang, Kaiwen Song, Xudong Li, Kerui Ren, Ran Yi, Lizhuang Ma, @chunhua_shen, @EvenEveno, @isaac__lt, Mulin Yu

tl;dr: anchor feed-forward visual priors to LiDAR–IMU motion for metric pose estimation and Gaussian reconstruction

https://t.co/PHsb3jaxtG

[MaxLiu09 @MaxLiu09]

https://t.co/uR1KcAoXYY

Made an open-source one that does the same thing, and there's a Blender extension available for it now too.

[Vincent Qin @AlphaRealcat]

@ShiboZhaoSLAM Hi Shibo, thanks for organizing this! Two quick questions: will the session recording be shared, and will the Top-10 teams' reports (and code) be released publicly? Would be a great resource for folks working on generalizable IMU odometry. Thanks!

[Shumpei Takezaki @shumpeiMaxwell]

Does the VGGT Family Need All Its Layers?

TL;DR: VGGT-style 3D models contain substantial layer redundancy. Systematic pruning reveals early/late removable regions, while simple linear calibration recovers accuracy—cutting aggregator parameters by up to 44% without end-to-end retraining.
https://t.co/iPh8nCs9hT

[Zhenjun Zhao @zhenjun_zhao]

LBDU-VIO: Learned Bias Dynamics and Uncertainty for Visual-Inertial Odometry with Unreliable Vision

Qizhi Guo, Junning Lyu, Defu Lin, Shaoming He

tl;dr: ODE model->continuous time bias dynamics; IMU uncertainty model->measurement noise covariance

https://t.co/lCfPcC3lKJ

[David Nordström @davnords]

Today we release Poincar3, our newest 3D model trained entirely through SSL on RGB image sequences.

I find the coolest thing is that, without labels, it learns multi-view geometry. Just using the attention map, reliable tracks can be formed (see below).
https://t.co/P1nShNSMR6

[Vincent Qin @AlphaRealcat]

@lilimamayu Awesome! 扫描效果如何，有demo吗？

[Vincent Qin @AlphaRealcat]

Homepage: https://t.co/kpwBY5mgjP

[Zhenjun Zhao @zhenjun_zhao]

MatcherCompass: A Deployment-Aware Benchmark to Guide Image Matcher Selection in the Wild

Hyunwoo Kim, @GiseopK

tl;dr: in title

https://t.co/lcMTNmzsZ0

[Vincent Qin @AlphaRealcat]

@pesarlin @eccvconf @mapo1 Congratulations!

[Zhenjun Zhao @zhenjun_zhao]

VI3: Grounding Pretrained 3D Foundation Models with Inertial Cues

Ernesto Lozano, Alberto Jaenal, @jcivera

tl;dr: IMU + 3D foundation model (3DFM)->metric 3DFM

https://t.co/oB2jXrSuuU

[Vincent Qin @AlphaRealcat]

@ShuoYangAIR @Kickstarter Cool robot!

[Vincent Qin @AlphaRealcat]

👏

[Radiance Fields @RadianceFields]

COLMAP 4.2 just released with LoMa, AMD acceleration, and a browser based viewer for COLMAP!

I'm most excited that spherical camera matching is also getting faster and more accurate!

[Minghan Qin @qin_minghan]

🚀 Introducing Lucida: composable Real-to-Sim scene modeling—turning real-world captures into composable 3D worlds.
Project Page: https://t.co/incxYK0EBG
HF: https://t.co/JNUKRJW7gU
Arxiv: https://t.co/mVgF6TQPpI

#Real2Sim #3DVision #EmbodiedAI #Robotics #ComputerVision #3DGen

[Jon Barron @jon_barron]

Here's my talk from the CVPR 2026 "Bitter Lessons" workshop earlier this summer. I've split it up into parts for the sake of discussion.

Part 1: in which I wax poetic about my youth and force the audience to look at my dissertation results.

[Zhenjun Zhao @zhenjun_zhao]

KLTNet: Learning Sparse Feature Tracking for Robust and Accurate Monocular Visual-Inertial Odometry

Renbiao Jin, Danping Zou, Wenxian Yu

tl;dr: global dense-flow initialization + triplet-patch refinement->learning-based KLT trackers

https://t.co/qvF4rDIZ30

[Vincent Qin @AlphaRealcat]

Code: https://t.co/Ig0YntCTlI

[Zhenjun Zhao @zhenjun_zhao]

RIPE++: Reinforced Keypoint Learning from Positive Pairs Only

Johannes Künzel, Peter Eisert, @HilsmannAnna

tl;dr: RIPE using positive pairs only; supervised with # RANSAC inlier+ # RANSAC outlier

https://t.co/CCYo3KoZzt

[Radiance Fields @RadianceFields]

4DGS from a single camera.

4DAnyone Create Anyone in 4D from a Casual Monocular Video (SIGGRAPH Asia)
Project: https://t.co/jvxJ7QkUpN
Code (Apache 2.0): https://t.co/MB2skadLQP
Model: https://t.co/6xSUSzdrw9