Baráth Dániel
@majti89
Tired of SLAM breaking in dynamic scenes? WildGS-SLAM (#CVPR2025) uses uncertainty-aware tracking & DINOv2-based maps to handle motion, remove dynamic objects, and enable high-quality view synthesis — all from monocular RGB. 👇 🌐wildgs-slam.github.io 📜arxiv.org/abs/2504.03886
🥳Excited to share our latest work, WildGS-SLAM: Monocular Gaussian Splatting SLAM in Dynamic Environments, accepted to #CVPR2025 🌐 We present a robust monocular RGB SLAM system that uses uncertainty-aware tracking and mapping to handle dynamic scenes.
I forgot the most important change: Now you can participate with 3rd party non-commercial models. DUSt3R, MASt3R, VGGSfM, VGGT, you name it. Let’s go!
#IMC2025 @CVPR challenge is online! This year: - back to collection_mapping: directory with many (similar) scenes->your job to reconstruct & cluster them - a bit more of training set - no transparent objects $50000 prize fund Deadline: June 2 kaggle.com/competitions/i… #IMW2025
Google Internship call: we are looking for a PhD student in the area of object/scene understanding and VLMs to join our Google team in Zurich next summer. If you have applied for the 2025 Google Internship Call already and are interested to know more, ping me!
SupeRANSAC: One RANSAC to Rule Them All @majti89 tl;dr: most accurate, but slow-ish RANSAC, by incorporating best practices & tricks: arxiv.org/abs/2506.04803 github.com/danini/superan… Sampling: PROSAC + P-NAPSAC for homography & rigid pose, PROSAC for epipolar 1/
SupeRANSAC: One RANSAC to Rule Them All @majti89 tl;dr: why RANSAC work well for different vision problems? ->answer->implementation details and problem-specific optimizations arxiv.org/abs/2506.04803
Excited to present our #CVPR2025 paper DepthSplat next week! DepthSplat is a feed-forward model that achieves high-quality Gaussian reconstruction and view synthesis in just 0.6 seconds. Looking forward to great conversations at the conference!
The use of AI in reviewing is a growing problem. Several of my ICCV papers have AI reviews -- one reviewer was so lazy that they left in the prompts! A common refrain that I hear is that people have difficulty writing in English and need to use AI to clean up their review.…
We are happy to share that this paper will be presented at CVPR 2025!
Learning Affine Correspondences by Integrating Geometric Constraints Pengju Sun, Banglei Guan, Zhenbao Yu, Yang Shang, Qifeng Yu, @majti89 tl;dr: DKM+AffNet? arxiv.org/abs/2504.04834
🏆 CrossOver is accepted as a 𝗛𝗶𝗴𝗵𝗹𝗶𝗴𝗵𝘁 at @CVPR #CVPR2025! ✨ 💻 Fully open-sourced code with all pre-trained checkpoints: github.com/GradientSpaces… 📡 Stay tuned for a deep-dive thread and what else we are cooking 🍳
🎉 Excited to share our latest work, CrossOver: 3D Scene Cross-Modal Alignment, accepted to #CVPR2025 🌐✨ We learn a unified, modality-agnostic embedding space, enabling seamless scene-level alignment across multiple modalities — no semantic annotations needed!🚀
Learning Affine Correspondences by Integrating Geometric Constraints Pengju Sun, Banglei Guan, Zhenbao Yu, Yang Shang, Qifeng Yu, @majti89 tl;dr: DKM+AffNet? arxiv.org/abs/2504.04834
How do you align point clouds, CAD, floor plans. images, and text to share scene knowledge? CrossOver learns a unified, modality-agnostic embedding space—enabling seamless multi-modal scene alignment without semantic labels. #CVPR2025 🌐sayands.github.io/crossover/
🎉 Excited to share our latest work, CrossOver: 3D Scene Cross-Modal Alignment, accepted to #CVPR2025 🌐✨ We learn a unified, modality-agnostic embedding space, enabling seamless scene-level alignment across multiple modalities — no semantic annotations needed!🚀
🔗 arXiv: arxiv.org/abs/2502.15011 📂 Project page: sayands.github.io/crossover/ Joint work with Ondrej Miksik, @mapo1, @majti89 and @ir0armeni 🤝💡
Just wrapped up an incredible time at KEPAF conference with the Hungarian image processing, machine learning, and computer vision community — great talks, fresh ideas, and inspiring discussions. Excited to see you next time! #KEPAF #ComputerVision #MachineLearning
Happy to share that this paper got accepted at WACV 2025 @wacv_official. The code is updated for easy use :) Feel free to check it out! Wish everyone a happy NY🥳 arxiv: arxiv.org/pdf/2406.16204 code: github.com/weitong8591/vop
Breaking the Frame: Image Retrieval by Visual Overlap Prediction @weitong8591, @PhilippCSE, @matas_jiri, @majti89 arxiv.org/pdf/2406.16204
#CVPR2025 kind reminder to reviewers : The review due date is Jan. 13. The review to author date is Jan. 23. Many thanks!🙏
Within 2025, I will be hiring several PhD candidates and Postdoctoral researchers @ICComputing to work on various aspects of #TopologicalDeepLearning (#TDL). Posts are attached. The postdoc positions are also available to apply under: imperial.ac.uk/jobs/search-jo… #CVPR #CVML #AI #ML
Niantic is hiring PhD students for internships in 2025: job-boards.greenhouse.io/niantic/jobs/7…
For people in Zurich, @FlorianCaesar is putting together ZurichCV #6 for free on November 19th. @majti89 from ETH Zurich on GLOMAP and @KarkusPeter from NVIDIA Research on Autonomous Driving in the Era of Foundation Models will present. RSVP Here: zurichai.ch/events/zurichc…
DepthSplat: Connecting Gaussian Splatting and Depth @haofeixu @songyoupeng @FangjinhuaWang @hermannsblum @majti89 @andreasgeiger0 @mapo1 tl;dr: infinite loop to train better unsupervised monodepth. arxiv.org/abs/2410.13862