#CVPR2026 reviewing -- this year my usefulness score is zero, meaning that my absence would not change any paper outcome.

There is something ironic in listening to Dance Monkey on repeat.

There is evidence that even robots with advanced AI capabilities prefer to assist their computer vision systems with manual finger-based attention, so probably visual attention and ViTs are a robot-evolutionary bottleneck.

After watching @veritasium.bsky.social videos about asbestos, Monsanto and PFAS I can understand why many people especially in US believe in conspiracy theories -- because there were extremely dangerous real conspiracies. P.S. Definitely recommend the videos.

RaCo: Ranking and Covariance for Practical Learned Keypoints Now on arXiv: arxiv.org/abs/2602.15755 AND code is there github.com/cvg/RaCo

🚀 Thrilled to share our survey paper: Advances in Global Solvers for 3D Vision The FIRST systematic survey unifying global optimization for 3D vision 📄 Paper: arxiv.org/abs/2602.14662 💻 Paper List & Tutorial Code : github.com/ericzzj1989/Awesome-Global-Solvers-for-3D-Vision 1/8

I see a lot of "Intel GPU" improvement in pytorch release blog, but no Apple. Is MPS device support dead?

LongStream: Long-Sequence Streaming Autoregressive Visual Geometry Chong Cheng, Xianda Chen, Tao Xie, Wei Yin, Weiqiang Ren, Qian Zhang, Xiaoyuang Guo, Hao Wang tl;dr: predict next frame relpose and metric scale separately, make KV cache to be same in test and train time. arxiv.org/abs/2602.13172

Nice! Finally ChatGPT started to recommend: - SuperPoint +SuperGlue - ratio test - mention that good RANSAC is important. Not yet recommended ALIKED, but confirmed it is great choice when asked chatgpt.com/share/6992cd...

Quick math question (@ccanonne.github.io?): let's say I sample n points iid from a gaussian in dimension d. What is the ratio of the volume of the biggest cone without samples, intersected with the smallest encompassing hypersphere, and that hypersphere. i.e., the relative volume without sample

TIL about Energy Vault: revolutionary startup, who proposes to store energy by lifting stones up. en.wikipedia.org/wiki/Energy_... P.S. Not so surprising, after the fact that boiling water is the way to use nuclear energy.

RAD: Retrieval-Augmented Monocular Metric Depth Estimation for Underrepresented Classes Michael Baltaxe, Dan Levi, Sagie Benaim tl;dr: use retrieval -> kNN -> matching at training and inference to get context how depth looks like based on the object class arxiv.org/abs/2602.09532

Revisiting [CLS] and Patch Token Interaction in Vision Transformers Alexis Marouani, Oriane Siméoni, Hervé Jégou, Piotr Bojanowski, Huy V. Vo tl;dr: ViT spends some capacity making sure CLS token is distinct. Using un-shared FFN for it may improve things. arxiv.org/abs/2602.08626

Raster2Seq: Polygon Sequence Generation for Floorplan Reconstruction Hao Phung, Hadar Averbuch-Elor tl;dr: predict a sequence of polygons (language), rather than corners and connect them. arxiv.org/abs/2602.09016 anonymous.4open.science/r/Raster2Seq...

Pixel-Accurate Epipolar Guided Matching Oleksii Nasypanyi, Francois Rameau tl;dr: if you do (for some reason) epipolar guided matching, this is probably the optimal way? Definitely improves upon our epipolar hashing from 2021 openreview.net/forum?id=9zR... #3DV2026

Understanding and Optimizing Attention-Based Sparse Matching for Diverse Local Features Qiang Wang tl;dr: what if you run LightGlue on descs it was trained on, with locations from other detector (R2D2, ORB, etc)? Results may surprise you IMC2021 eval. NMS is must. arxiv.org/abs/2602.08430

Homography Decomposition Revisited Yaqing Ding, Jian Yang, Zuzana Kukelova tl;dr: faster than SVD decomposition of H to (R,t) or E, based on using 3 pt correspondences. link.springer.com/article/10.1...

Understanding and Optimizing Attention-Based Sparse Matching for Diverse Local Features Qiang Wang tl;dr: detector matters more, not descriptor! removing nearby keypoints matters! decouple det & desc; fine-tuning detector-agnostic eval on IMC2021 arxiv.org/abs/2602.08430

Marginalized Bundle Adjustment: Multi-View Camera Pose from Monocular Depth Estimates Shengjie Zhu, Ahmed Abdelkader, Mark J. Matthews, Xiaoming Liu, Wen-Sheng Chu tl;dr: BA for monodepth. IMC2021 results! openreview.net/forum?id=OMT...

Fundamental Constraints on Camera Centers Arising from 3D-2D Matches of Points and Lines Jaired Collins, Taci Ata Kucukpinar, Timothy Duff, Joshua Fraser, Guna Seetharaman & Kannappan Palaniappan tl;dr: Results for(1point + 1 line) and (2 lines ) link.springer.com/content/pdf/...

RaCo: Ranking and Covariance for Practical Learned Keypoints Abhiram Shenoi, Philipp Lindenberger @pesarlin.bsky.social @marcpollefeys.bsky.social tl;dr: ALIKED arch + DaD RL train+full 360rotaug. Det+covariance heads, separate ranker model. No IMC eval #3DV2026 openreview.net/forum?id=BWt...

The Illusion of Generalization: Re-examining Tabular Language Model Evaluation Aditya Gorla, Ratish Puduppully tl;dr: @tunguz.bsky.social and other Kaggle GMs were right all along about tabular DL models. arxiv.org/abs/2602.04031

RayRoPE: Projective Ray Positional Encoding for Multi-view Attention Yu Wu, Minsik Jeon, Jen-Hao Rick Chang, @onceltuzel.bsky.social Shubham Tulsiani tl;dr: even more projective geometry based PE, if you know your camera pose. arxiv.org/abs/2601.15275

S-MUSt3R: Sliding Multi-view 3D Reconstruction Leonid Antsfeld, Boris Chidlovskii, Yohann Cabon, @vincentleroy.bsky.social Jerome Revaud tl;dr: sliding MUSt3R for cheap long seq The most interesting is alignment: point-based+camera-based. KDTree local desc for loop closure arxiv.org/abs/2602.04517

CoWTracker: Tracking by Warping instead of Correlation Zihang Lai, Eldar Insafutdinov, Edgar Sucar, Andrea Vedaldi tl;dr: warping head does help, so as using pretrained VGGT as a backbone. arxiv.org/abs/2602.04877

Wid3R: Wide Field-of-View 3D Reconstruction via Camera Model Conditioning Dongki Jung, Jaehoon Choi, Adil Qureshi, Somi Jeong, Dinesh Manocha, Suyong Yeon tl;dr: Pi-3 with (ray+radial distance) point map prediction and wide/fisheye camera tokens. arxiv.org/abs/2602.05321

Predicting Camera Pose from Perspective Descriptions for Spatial Reasoning Xuejun Zhang, Aditi Tiwari, Zhenhailong Wang, Heng Ji tl;dr: injecting camera poses info into LLM to help answer 3D related questions arxiv.org/abs/2602.06041

S-MUSt3R: Sliding Multi-view 3D Reconstruction Leonid Antsfeld, Boris Chidlovskii, Yohann Cabon, @vincentleroy.bsky.social, Jerome Revaud tl;dr: sliding-window MUSt3R; overlapped input segments->MUSt3R->loop closure->PGO with iterative solver arxiv.org/abs/2602.04517

CoWTracker: Tracking by Warping instead of Correlation Zihang Lai, Eldar Insafutdinov, Edgar Sucar, Andrea Vedaldi tl;dr: iterative warping on current estimation,then spatio-temporal transformer refines tracks arxiv.org/abs/2602.04877

🎉 Thrilled to share our ICLR 2026 paper: 🔹 NPC Neural Predictor-Corrector: Solving Homotopy Problems with Reinforcement Learning 🚀 The first unified framework that reveals diverse problems share a common predictor-corrector structure 📄 Paper: arxiv.org/abs/2602.03086 💻 Code: [Coming soon] 1/

Incrementing your names can do wonders mp3 -- just sound mp4 -> now you have video! mp5 - now you have a machine gun!

BBoxMaskPose v2: Expanding Mutual Conditioning to 3D Miroslav Purkrabek, Constantin Kolomiiets, Jiri Matas tl;dr: sota in human pose estimations, especially for the hard cases arxiv.org/abs/2601.15200

CzechLynx: A Dataset for Individual Identification and Pose Estimation of the Eurasian Lynx Lukáš Picek et 16 al. tl;dr; pose estimation, but for Lynx. 39k labeled images from 2009 to 2025. Synthetic renders as a bonus. arxiv.org/abs/2506.04931

XRefine: Attention-Guided Keypoint Match Refinement Jan Fabian Schmid, Annika Hagemann tl;dr: pairwise kpt2subpix with self-attention and study how kpt accuracy influences the camera pose accuracy arxiv.org/abs/2601.12530

The new CTU Rector begins their term in office with strong support for excellence. CTU has just launched a Starting Grant to attract outstanding early‑career researchers who wish to join CTU and establish their own research group. Funding: up to €160k per year for 3 years. Deadline: 30 March 2026.

I have nothing personal with Walter. On the contrary, I honestly thank him for all his service for science! The following is a complaint for our community as a whole. This PAMI-TC newsletter clearly shows that we are doing business as usual, as if nothing is happening in the world right now. 1/3

Modern monetization strategy of mobile games is evil. Remember classic Lemmings? It was an amazing 1991 game. Later, in 2018 there was AMAZING Android port Revisited Lemmings. It was free, but I am fine paying $10 for it. No! Today you have - spinwheels, limited lives trash. And design is bad.

Recently I have bought, but haven't really read (just chapter 1) the Source Code book. I guess I just toss it away now?

Is there any current 3/4D reconstruction method able to accurately reconstruct this scene? (with correct size of thunderstorm)

The #ECCV2026 Malmo 🇸🇪 call for papers is now available. Check it out! Call for Papers: eccv.ecva.net/Conferences/...

Vega plots and confusion tables in wandb are such a useless scam :( Like, who needs to have "summary" confusion table, I need it per val step. matplotlib -> log as image, here we go again.

If Bilbo had Chat GPT

VGGT-SLAM 2.0: Real time Dense Feed-forward Scene Reconstruction Dominic Maggio, @lucacarlone.bsky.social tl;dr: no 15 DoF alignment in VGGT-SLAM; keyframes as nodes+intra-/inter-submap edges->factor graph; attention layers in VGGT->image retrieval verification arxiv.org/abs/2601.19887

ChatGPT has a sense of humor.

git commit -m "why am I so stupid"

Quick update: there was an access issue with the DAD-3DHeads dataset, but it’s now fixed ✅ The dataset is available now on Hugging Face: 👉 huggingface.co/datasets/t-m... P.S. There is no need to send us your test submissions for evaluation anymore, you can do it yourselves now! You're welcome :)

I am reading now the reviews by fellow reviewers (I didn't submit for CVPR) in my batch, and, OMG, I see so many ChatGPT reviews.

still, google scholar does the shortest exports out of all (fairly easy to check & minimize further for the paper). compare that with bib entries that my students find: some go to give you the street name of the conference venue and the names of the parents of the program chairs 🙃

The day has arrived! #CVPR2026 reviews are scheduled to be released sometime today. NO, we don’t know the exact release time, we’ll find out together with the community. Only our Program Chairs do. Good luck to everyone! 🤞

XRefine: Attention-Guided Keypoint Match Refinement Jan Fabian Schmid, Annika Hagemann tl;dr: Keypt2Subpx with cross-attention w/o descriptor input; extend to multi-view setting no eval on IMC arxiv.org/abs/2601.12530