iiinova — where your digital self keeps evolving

David Nordström @davnords.bsky.social · 13/09/2026

New matcher unlocked. Also some VGGT-interp stuff

David Nordström @davnords.bsky.social · 02/06/2026

Heading to #CVPR26 now! You'll find me at the Image Matching Worskhop on June 4th talking about rotation invariant matching and on June 6th presenting MuM as a poster. If you are in Denver, let's meet up!

120

Reposted by David Nordström

Dmytro Mishkin @ducha-aiki.bsky.social · 29/05/2026

Image matching since 2020: 2020: @pesarlin.bsky.social SuperGlue 2023: @vincentleroy.bsky.social DUSt3R 2024: @parskatt.bsky.social RoMa 2025: @jianyuanwang.bsky.social VGGT 2026: @davnords.bsky.social"hold my beer" scales LightGlue) 2026: @jianyuanwang.bsky.social "no, hold MY beer"(scales VGGT)

0225

Reposted by David Nordström

Dmytro Mishkin @ducha-aiki.bsky.social · 17/04/2026

It seems, that we have failed the communication about IMC26. Let's try again. The competition this year is here: kaggle.com/competitions... No prizes, but whole year leaderboard -- similar to KITTY and other academic competitions. 3D people, please retweet and share.

kaggle.com

Image Matching Challenge 2025 Ongoing

Ongoing leaderboard for Image Matching Challenge 2025.

0137

David Nordström @davnords.bsky.social · 14/04/2026

LoMa-R is our newest addition to the LoMa family. In our IMW paper at #CVPR26, we investigate rotation invariance in the sparse matching pipeline. The resulting model is robust to rotations, even matching star constellations, and achieves strong upright performance. github.com/davnords/loma

github.com

GitHub - davnords/LoMa: LoMa: Local Feature Matching Revisited

LoMa: Local Feature Matching Revisited. Contribute to davnords/LoMa development by creating an account on GitHub.

3101

Reposted by David Nordström

Georg Bökman @bokmangeorg.bsky.social · 14/04/2026

Sparse image matching is done via 1) keypoint detection in each image, 2) keypoint description, 3) matching of descriptions between images. Should rotation invariance be enforced at stage 2 or 3? Turns out both work fine! To be presented at the CVPR image matching workshop by @davnords.bsky.social

0102

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 14/04/2026

Accepted to the Image Matching Workshop at #CVPR26!

1171

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 14/04/2026

Introducing LoMa-R! A rotation invariant version of LoMa. Code: github.com/davnords/loma Paper: arxiv.org/abs/2604.11809

1183

Reposted by David Nordström

Dmytro Mishkin @ducha-aiki.bsky.social · 07/04/2026

LoMa: Local Feature Matching Revisited @davnords.bsky.social @parskatt.bsky.social , @bokmangeorg.bsky.social et 6 al tl;dr: if you train DeDoDe+LightGlue on VGGT-scale data, it helps a LOT. New IMC2022 sota arxiv.org/abs/2604.04931

1142

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 07/04/2026

Introducing LoMa, the next generation of feature matcher!

3414

David Nordström @davnords.bsky.social · 13/03/2026

Yesterday, @parskatt.bsky.social and the RoMa v2 team got an award at the annual Swedish Symposium on Image Analysis. FeelsGoodMan. @bokmangeorg.bsky.social @fredkahl.bsky.social @jastermark.bsky.social @liucvl.bsky.social et. al.

3143

David Nordström @davnords.bsky.social · 23/02/2026

Biggest psyop: Learning rate decay. @parskatt.bsky.social showed me the truth but I refused to accept it...

260

Reposted by David Nordström

Yaroslava Lochman @ylochman.bsky.social · 25/11/2025

Presenting today at #BMVC2025 our follow-up work on anisotropic rotation averaging which is particularly useful in global SfM. We propose a fast solver ACD and integrate robust optimization. If you’re at the conference, welcome to come to our poster #516! bmvc2025.bmva.org/proceedings/...

072

David Nordström @davnords.bsky.social · 24/11/2025

We are introducing MuM, a feature encoder (ViT-L) tailored for 3D vision tasks. TLDR; Spiritual successor to CroCo with a simpler multi-view objective and larger scale. Beats DINOv3 and CroCo v2 in RoMa, feedforward reconstruction, and rel. pose. arxiv.org/abs/2511.17309 github.com/davnords/mum

3394

David Nordström @davnords.bsky.social · 20/11/2025

Johan out here carrying Swedish CV academia with a 10 author absolute masterclass. Only the avatar can master all four elements (LiU, CTH, LTH and UvA). Very strong and versatile model overall. RoMa v2 can match anything you'd like. Even @bokmangeorg.bsky.social 's hand annotated satellite images.

180

Reposted by David Nordström

Dmytro Mishkin @ducha-aiki.bsky.social · 20/11/2025

RoMa v2: Harder Better Faster Denser Feature Matching @parskatt.bsky.social et 11 al. tl;dr: in title. Predict covariance per-pixel, more datasets, use DINOv3, adjust architecture. arxiv.org/abs/2511.15706

3142

Reposted by David Nordström

Georg Bökman @bokmangeorg.bsky.social · 21/10/2025

Turns out NLP is just vision

2235

David Nordström @davnords.bsky.social · 22/10/2025

These gentlemen show how not only colmap but also VGGT fail on spherical motion in their ICCV oral paper "Uncalibrated Structure from Motion on a Sphere". I wonder if it is just a data issue for VGGT or if it is deeper than that. I mean VGGT was trained on mostly synthetic data. What do you think?

1101

Reposted by David Nordström

Georg Bökman @bokmangeorg.bsky.social · 16/10/2025

Pro tip: For good Halloween vibes, use non-normalized RoPE on images larger than your training resolution and larger than the composite period of some of the RoPE-rotations. You might get scary ghost structures in your features.

1113

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 23/09/2025

RoMa now on PyPI under name of `romatch`

172

Reposted by David Nordström

Dmytro Mishkin @ducha-aiki.bsky.social · 18/09/2025

Towards the Next Generation of 3D Reconstruction @parskatt.bsky.social PhD Thesis. tl;dr: would be useful in teaching image matching - nice explanations. (too) Fancy and stylish notation. Cool Ack section and cover image. liu.diva-portal.org/smash/record...

2338

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 17/09/2025

And here is a link to the thesis itself: liu.diva-portal.org/smash/record...

liu.diva-portal.org

Towards the Next Generation of 3D Reconstruction

0196

Reposted by David Nordström

Christian Wolf @chriswolfvision.bsky.social · 13/09/2025

How to name your method: a comprehensive flow chart

14310

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 26/07/2025

Born too late to explore the earth. Born too early to explore the galaxy. Born just in time to \nabla_{\theta}f_{\theta}

1101

Reposted by David Nordström

Georg Bökman @bokmangeorg.bsky.social · 15/07/2025

Tomorrow at ICML [Tuesday 4:30 pm, poster W-213] @davnords.bsky.social and I will present our spotlighted flop paper. Come by and let us try to convince you that equivariant nets should be standard in vision tasks due to computational benefits! bsky.app/profile/bokm...

Poster for the paper "Flopping for FLOPs".

0163

David Nordström @davnords.bsky.social · 14/07/2025

I am at ICML, happy to connect with all of you :)

000

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 13/06/2025

And today I'm presenting this work at #CVPR2025! 🗓️ Date: 16:00-18:00, Fri, Jun 13 (Today) 📍Place: Poster #115 in Session 2 (ExHall D) 💻 Code: github.com/ericssonrese...

0142

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 23/05/2025

As models become larger, more of their compute is spent in the MLP. Turns out that this is perfect for octic equivariance, as our biggest gains are there!

161

David Nordström @davnords.bsky.social · 23/05/2025

Want stronger Vision Transformers? Use octic-equivariant layers (arxiv.org/abs/2505.15441). TLDR; We extend @bokmangeorg.bsky.social's reflection-equivariant ViTs to the (octic) group of 90-degree rotations and reflections and... it just works... (DINOv2+DeiT) Code: github.com/davnords/octic-vits

2294

David Nordström @davnords.bsky.social · 28/03/2025

GPT 4o's new image capabilities seem to be liked. The insinuation from OpenAI seems to be that it is not based on diffusion. I wonder how their work relates to the infamous NeurIPS paper "Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction" (arxiv.org/abs/2404.02905)

130

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 24/03/2025

New paper! We merge SfM reconstructions with point cloud registration. Link: arxiv.org/abs/2503.17093 Code: Not yet public, but coming later.

9518

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 18/03/2025

New paper! (arxiv.org/abs/2503.13433), we look into improving the threshold roubustness of Random Sample Consensus (RANSAC) through (less biased) inlier noise scale estimation.

4308

Reposted by David Nordström

Jianyuan Wang @jianyuanwang.bsky.social · 17/03/2025

Introducing VGGT (CVPR'25), a feedforward Transformer that directly infers all key 3D attributes from one, a few, or hundreds of images, in seconds! Project Page: vgg-t.github.io Code & Weights: github.com/facebookrese...

34514

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 11/03/2025

Introducing DaD, Part 2, a pretty cool keypoint detector.

5315

Reposted by David Nordström

Johan Edstedt @parskatt.bsky.social · 10/03/2025

We made a new keypoint detector named DaD, paper isn't up yet, but code and weights are: github.com/Parskatt/dad

7448

Reposted by David Nordström

Georg Bökman @bokmangeorg.bsky.social · 10/02/2025

Common beliefs about equivariant networks for image input include 1) They are slow. 2) They don’t scale to ImageNet. 3) They are complicated. In my opinion, these three are all false. To argue against them, we made minimal modifications to popular vision models, turning them mirror-equivariant.

2264