# Interesting papers ## VGGT-SLAM: Dense RGB SLAM Optimized on the SL(4) Manifold - VGGT: https://vgg-t.github.io/ - incrementally and globally aligning submaps created from the feed-forward scene reconstruction approach VGGT - VGGT에서 GPU mem 부족으로 리컨할 수 없던 것을 리컨 가능하게 만듬 - Uncalibrated 카메라로 reconstruction을 할 때에는 scale을 알 수 없는데, Special Linear Group of degree 4 (SL(4))를 사용하면 된다? <img width="1399" alt="Image" src="https://github.com/user-attachments/assets/9e9a198c-ee53-40dc-ae8a-69b463f72571" /> <img width="1399" alt="Image" src="https://github.com/user-attachments/assets/625fe581-3287-4632-a4d1-78ec30265c9b" /> <img width="1399" alt="Image" src="https://github.com/user-attachments/assets/e5007f6b-411b-4501-8775-2fd4a9f93ef8" /> ## ASHiTA: Automatic Scene-grounded HIerarchical Task Analysis - 로봇은 '진짜' high-level task를 풀 수 있는가? <img width="611" alt="Image" src="https://github.com/user-attachments/assets/ef4a2f01-5d87-46f5-be7f-15b608d3cf07" /> <img width="1191" alt="Image" src="https://github.com/user-attachments/assets/249a3312-089b-488b-a947-c8c3be97772c" /> <img width="597" alt="Image" src="https://github.com/user-attachments/assets/2ffcbb0c-0569-44b6-9600-8d70aebfb169" /> <img width="597" alt="Image" src="https://github.com/user-attachments/assets/a0d42f9f-3842-4da4-a68c-3650d5b9cd85" /> # Google I/O 2025 - XR Glasses! 삼성 (젠틀몬스터), XReal 협업  