Can Gümeli

Check out ObjectMatch, our new #CVPR2023 paper on semantic registration! We use objects to register frames w/o view overlap.

@ducha_aiki @ChuanxiaZ @zhang_ganlin Cool work @wrchen530 !

NOVA3R: Non-pixel-aligned Visual Transformer for Amodal 3D Reconstruction Weirong Chen @ChuanxiaZ @zhang_ganlin Andrea Vedaldi, Daniel Cremers tl;dr: let VGGT output latents -> decode point cloud. arxiv.org/abs/2603.04179

@chrisoffner3d It was also true for .log/exp etc., hope numpy people fix that

np.mean(x) ✅ x.mean() ✅ np.median(x) ✅ x.median() 🔥😱💀 Whyyy?

@ducha_aiki I agree with “Yes” part only without details…

Do you trust RANSAC stopping criterion?

📢 LiteReality: Graphics-Ready 3D Scene Reconstruction from RGB-D Scans🏠✨ -> converts RGB-D scans into compact, realistic, and interactive 3D scenes — featuring high-quality meshes, PBR materials, and articulated objects. 📷youtu.be/ecK9m3LXg2c 🌍litereality.github.io

PrEditor3D enables smooth editing of the regions you want to change in a 3D shape, while preserving the rest. If you’re interested in precise 3D shape editing with a simple user interface, check out our poster this morning!

📢📢 We’ll be presenting MeshArt tomorrow morning (Friday 13.06) in the poster session at ExHall D Poster #42 from 10:30-12:30. Come and chat about articulated 3D mesh genereation or any 3D generative stuff! Project page: daoyig.github.io/Mesh_Art/

Presenting PrEditor3D at #CVPR2025 📢📢 If you'd like to learn more about our work and discuss 3D generation/editing, come visit our poster on Friday, June 13th, in ExHall D between 10:30-12:30 (Poster #44). Project Page: ziyaerkoc.com/preditor3d

📢PBR-SR: Mesh PBR Texture Super Resolution from 2D Image Priors📢 We propose a new optimization to up-sample textures of 3D assets (albedo, roughness, metallic, and normal maps) by leveraging 2D super-resolution models. 📝arxiv.org/abs/2506.02846 📽️youtu.be/eaM5S3Mt1RM

🚀"Leveraging Automatic CAD Annotations for Supervised Learning in 3D Scene Understanding" We provide high-quality CAD model and pose annotations for the ScanNet++ v1 dataset created with SCANnotate++ method. 🔗 Project Page: stefan-ainetter.github.io/SCANnotatepp/ #3DVision #ComputerVision

📢 QuickSplat: Fast 3D Surface Reconstruction via Learned Gaussian Initialization @liuyuehcheng learns 2DGS initialization, densification, and optimization priors from ScanNet++ => fast & accurate reconstruction! Project: liu115.github.io/quicksplat

Since March 19, I have been unjustly imprisoned. Since my arrest, millions of Turks of all walks of life raised their voices for justice and democratic rights. Thousands beyond our borders joined this dignified stance in a global gesture of solidarity. Governments of France and Germany, mayors of over 75 European cities, 24 international city networks including the B40 which I founded, the four major political groups in the European Parliament, the United Nations bodies, the Council of Europe, Amnesty International, Freedom House, and Human Rights Watch—some of the world’s most respected institutions—have spoken out firmly against this injustice. Paris granted me honorary citizenship. City halls in Bologna, Florence, and Modena displayed banners calling for my freedom. Mayors and representatives of 20 German cities came to Istanbul to express their support. Diplomats from the U.S., Germany, France and over 10 countries monitored my hearings. The European Parliament and Council of Europe decided to send fact-finding missions to Istanbul. I know we, the Turkish people, will prevail because just causes always triumph in the end. #SolidarityWins #JusticePrevails

📢Pixel3DMM: Versatile Screen-Space Priors for Single-Image 3D Face Reconstruction📢 -> highly accurate face reconstruction by training powerful VITs via surface normals and UV-coordinates estimation. The geometric cues from our 2D foundation model backbone constrain the 3DMM parameters, which allows us to achieve remarkable reconstruction accuracy - works for both single image and videos! In addition, we introduce a new 3D face reconstruction benchmark that evaluates both neutral and posed face geometry. 🌍 simongiebenhain.github.io/pixel3dmm 📷 youtu.be/BwxwEXJwUDc Great work by @SGiebenhain @TobiasKirschst1 @martin_ruenz @LourdesAgapito

Leveraging Automatic CAD Annotations for Supervised Learning in 3D Scene Understanding @YuchenRao Stefan Ainetter, @stekovic_sinisa Vincent Lepetit, Friedrich Fraundorfer tl;dr: in title arxiv.org/abs/2504.13580

📢 IntrinsiX: High-Quality PBR Generation using Image Priors 📢 From text input, we generate renderable PBR maps! Next to editable image generation, our predictions can be distilled into room-scale scenes using SDS for large-scale PBR texture generation. We first train separate LoRA modules for the intrinsic properties of albedo, rough/metal, normal. Then, we introduce cross-intrinsic attention using a rerendering loss with importance-weighted light sampling to enable coherent PBR generation. Our method outperforms text -> image -> PBR methods both in generalization and quality, since directly generating PBR maps does not suffer from the inherent ambiguity of intrinsic image decomposition. In addition, our design choice facilitates SDS-based PBR texture distillation. 🌍 peter-kocsis.github.io/IntrinsiX/ 🎥 youtu.be/b0wVA44R93Y Great work by @Peter4AI, @LukasHollein

📢ExCap3D: Multilevel Captioning of Objects in 3D Scenes @chandan__yes generates consistent object and part-level descriptions of objects in 3D scenes, and introduces a new dataset with 190k captions for 34k ScanNet++ objects. Project: cy94.github.io/excap3d w/ @david_roz_

📢Announcing our 3D head avatar benchmark📢 Two tasks with hidden test sets: - Dynamic Novel View Synthesis on Heads - Monocular FLAME-driven Head Avatar Reconstruction Our goal is to make research on 3D head avatars more comparable and ultimately increase the realism of digital humans. The benchmark studies distinct phenomena of 3D head avatar creation, such as extreme facial expressions, slow motion captures of shaking long hair, or complicated light reflection and refraction patterns of glasses. The two benchmark tasks assess two core desiderata of 3D avatars: While the novel view synthesis challenge focuses on best possible rendering quality of complex moving scenes, the avatar animation challenge is concerned with how well a driving signal is translated into an avatar. Evaluations are light-weight and consist of diverse video recordings from the popular NeRSemble dataset with a hidden test set. Participation in the benchmark is therefore straight-forward and requires only 5 reconstructions per task. Leaderboard and benchmark submission: kaldir.vc.in.tum.de/nersemble_benc… Benchmark data access and toolkit: github.com/tobias-kirschs… Great work by @TobiasKirschst1 @SGiebenhain

In the early hours of March 19, I was detained by dozens of armed police officers. It was a political operation — not a legal one — timed just days before CHP’s presidential primary. But what followed was far more powerful than the repression itself: people from all across Türkiye stood up. From Istanbul to Rize, citizens took to the streets, joined our party, and reminded us all that democracy lives in the will of the people. As I wrote in my op-ed for The New York Times, “I have faith in the people of Türkiye and beyond who fight for justice and democracy.” This struggle is not just about one candidacy or one city — it is about rebuilding a republic that belongs to its citizens, not to fear. nytimes.com/2025/03/28/opi…

Turkish democracy was ailing even before the main contender for challenging (and perhaps winning against) President Erdoğan, the Istanbul mayor Ekrem İmamoğlu was arrested last week on charges that many are arguing to be trumped up. This by itself is noteworthy. It has become more so because it has also triggered widespread protests, which Turkey has not experienced since those surrounding the Gezi Park were put down. What makes all of this so significant is that the current geopolitical situation has created a new opportunity for Turkey, which the country should not squander. It goes something like this. Trump is attempting to remake the world order, not just fundamentally damaging American democracy. The US is increasingly allying itself with nondemocratic, authoritarian countries. The president is famously keen on staying in good terms with Russia’s autocratic leader Vladimir Putin. Despite all the talk of tariffs and chip wars, Trump could also form a new, and more friendly, transactional relationship with China. The countries that Trump seems less interested in aligning the US with are the European ones and Canada. He’s also signaling the possibility of the US reneging its commitments to NATO (something that would have been unthinkable just a few years ago). This could all change, if Democrats take back the House of Representatives in the midterms in 2026 and slow down Trump’s attack on US institutions and efforts to remake the global order, and then when the 2028 presidential election to reverse their consequences and rebuild US democracy. But so far, there is no guarantee that they will be able to do so. We should certainly not count on US institutions surviving intact for four more years. The possibility of a new European block (perhaps joined by Canada), committed to liberal democracy and its own defense, is no longer unthinkable. In the context of these developments, Turkey will have a choice. It can remain outside of both blocks. It can cozy up to the US-Russia-China axis. Or it can become part of the European block. The European path is not a pie-in-the-sky. Turkey already has the second-largest ministry in NATO, and would be a crucial partner to Germany, France and other European countries in building a common defense. With rapid aging of European populations, the specter of more Turkish immigration to Europe may no longer be as daunting. Becoming part of the European Union and the European defense pact would be game changers for Turkey. The country’s economy has been struggling because of lackluster productivity growth since 2006, and consequently slow real wage growth, despite endemic poverty. European markets, European capital, and most importantly European technology and joint venture opportunities with European companies could substantially boost Turkey’s productivity potential. (Recall, in this context that, the most rapid period of productivity growth in Turkey's recent past, between 2002-2006, was in the context of the country’s EU accession talks). Yet, this path requires Turkey’s commitment to democracy. Turkey needs to credibly signal to its European partners that it will try to reverse the downward slide of its democratic institutions (which has given the country the dubious achievement of being one of the worst performers, globally, in terms of worsening of democracy). It will require its government to accept and even celebrate civil society activity and protests. It will require the country’s youth to build on its newfound, greatest enthusiasm for politics and become even more engaged in the country’s affairs. It will require the population to see Europeans as partners (that of the open sequence nationalist fervor directed against all Western powers). Who will take this path? President Erdoğan has proven over the last several decades that he can change with the times and opportunities (and can bring his base together when he makes such turnarounds). But the change in perspective and institutional approach that is required here may be more radical than trying to bring the Kurds into the fold (which the president is now attempting). There is no guarantee that Turkey will take this path. If President Erdoğan does not attempt it (as is likely), Turkish politics will become more uncertain and more confrontational. It remains unclear who will take the lead and whether this leader will be allowed to kickstart the necessary epochal transformation for the country’s future.

Gotta catch ’em all? Tell that to the Pikachu who just outran the Turkish police in Antalya.