Patrick Wenzel

"Building Rome with Convex Optimization" has been accepted to #RSS2025! Try XM, our new structure from motion pipeline powered by GPU-accelerated convex semidefinite optimization: github.com/ComputationalR… XM solves large-scale (nonconvex) global bundle adjustment problem via learned depth and a tight convex semidefinite relaxation. By implementing the Burer-Monteiro low-rank factorization algorithm in CUDA, XM can solve bundle adjustment problems with more than 10,000 images/views. Technical details in the paper: arxiv.org/abs/2502.04640 Kudos to @Hyhan0118

Unanswered research questions in Image Matching

One reason generative 3D is hard is that the world has barely any 3D training data. Today, most humans on earth will snap a few photos and write paragraphs of text. In contrast, earth has maybe ~50k 3D artists, and each will make maybe ~1k photoreal 3D models over their career.

We’ve had fun testing the limits of MASt3R-SLAM on in-the-wild videos. Here’s the drone video of a Minnesota bowling alley that we’ve always wanted to reconstruct! Different scene scales, dynamic objects, specular surfaces, and fast motion.

Introducing MASt3R-SLAM, the first real-time monocular dense SLAM with MASt3R as a foundation. Easy to use like DUSt3R/MASt3R, from an uncalibrated RGB video it recovers accurate, globally consistent poses & a dense map. With @eric_dexheimer*, @AjdDavison (*Equal Contribution)

Devil's advocate mode on: Navigation World Models have existed for a long time... they're called maps! And there are plenty of good algorithms out there which enable robots to build them / render views from them / localise within them / use them for planning. #SLAM #SpatialAI :)

I’ll be at the conference tomorrow and on Friday. PM me if you’d like to chat about what we are doing!

Mark Zuckerberg on Meta's open source AI strategy: "We're not doing this because we're altruistic... I don't view it as giving it away, I view it as you guys all making it better for me"

very cool to see a pretty exhaustive and extensive technical report on llama-3.1! a few fun snippets 🧵

🚀 Excited to release OpenVPRLab! 🎉 An open-source framework for Visual Place Recognition (VPR), featuring extensible, modular, and scalable components, enabling researchers to train/develop deep VPR models with reproducible SOTA performance. 🔗github.com/amaralibey/Ope… 🧵👇

DMs are open now!

We’re looking for computer vision and machine learning PhD interns to join our team. It’s a great opportunity to work on challenging and cutting-edge computer vision and machine learning problems! I’ll be at #CVPR this week. PM me! linkedin.com/jobs/view/3953…

What a great community. We had a full house yesterday and a great lineup of keynote speakers! We had stimulating discussions and a really exciting panel! We hope everyone enjoyed it! Thanks to all the speakers and to the organizers who contributed to the success of the event!

@valzadrija @CVPR @smash0190 @lukasvst @VincentLepetit2 @jajuengel @lealtaixe @mapo1 The recording will be available on YouTube after approximately 3 months.

@PatrickWenzelML @CVPR @smash0190 @lukasvst @VincentLepetit2 @jajuengel @lealtaixe @mapo1 Is there going to be YouTube streaming? Thanks in advance 🙏

Make sure to join us for the Visual Localization and Mapping Workshop (ViLMa) on Monday morning @CVPR! We have an outstanding lineup of keynote speakers: @smash0190 @lukasvst @VincentLepetit2 @jajuengel Peter Kontschieder @lealtaixe @mapo1 More details: vilma-workshop.github.io

With satellite imagery, it’s hard to get labels. Solution? DINOv2! WRI+Meta trained a satellite DINOv2 for tree height estimation. They created an interactive map of tree height of the whole globe (!) at 1-meter res (!): …toring-okw37.projects.earthengine.app/view/canopyhei… Quizz: Can you recognize this city?

Lecture recordings from MIT about inverse rendering. Teaches also some Gaussian Splatting basics for free! scenerepresentations.org/courses/invers… Credits go to @cs_mshah for bringing it to my attention #3DGS

The GOAT of tennis @DjokerNole said: "35 is the new 25.” I say: “60 is the new 35.” AI research has kept me strong and healthy. AI could work wonders for you, too!

Super excited to announce that our workshop ViLMa - Visual Localization and Mapping has been accepted at @CVPR. Please stay tuned for more updates & see you all in Seattle!

Helsing is hiring 2024 PhD interns in the area of 3D computer vision, machine learning, and generative AI. See intern call below for more details. helsing.ai/jobs/4264969101