우리는 영상을 생성하지 않는다. 영상으로부터 세계를 생성한다. 실시간 4D 월드 모델 'InSpatio-World' 공개 — 입력 비디오 클립을 동적·탐색 가능한 지속적 세계로 변환. 시점 자유 탐색과 시간의 앞뒤 제어 가능. 라이브 데모, 코드·가중치, 프로젝트 페이지 제공.
We don’t generate videos. 🎬 We generate worlds from videos. 🌍 Introducing InSpatio-World — the world's first open-source real-time 4D world model‼️ Your input: a video clip Our output: a dynamic, navigable, persistent world 🕹️ explore freely across viewpoints ⏪ control
XGRIDS (@XGRIDS_OFFICIAL)
AWS 부스(#921)에서 Real2Sim 파이프라인 시연 예정이라는 안내 트윗입니다. 모바일로 스캔한 실제 공간을 고해상도 3D 월드로 변환해 로봇 트레이닝에 활용하는 워크플로우(Real2Sim)를 GTC2026 행사에서 3–5PM에 시연한다고 홍보합니다. Spatial AI 및 로보틱스용 시뮬·데이터 생성의 실무 적용 사례로 주목됩니다.
Rohan Paul (@rohanpaul_ai)
OpenClaw은 로봇용 실시간 3D 추론 시스템으로 Unitree G1에서 동작한다. 동적 복셀 맵을 생성해 객체와 시간적 변화를 추적하고, 임베딩된 공간 표현으로 복잡한 시공간 질의(딥 스페이셜 쿼리)를 지원한다. 드론과 사족보행 로봇(퀘드러페드)에 적합한 지각 솔루션으로 로봇 인식 및 내비게이션 향상을 목표로 한다.
🆕 If someone says “world model,” ask: which kind?
✅ From tokens to environments: what “world model” is trying to mean
→ https://gradientflow.substack.com/p/world-model-is-a-mess-heres-how-to
#AI #ArtificialIntelligence #GenAI #MachineLearning #DeepLearning #WorldModels #SpatialAI #EmbodiedAI #PhysicalAI #Robotics #ComputerVision
🧰 Practical GeoAI Tools
Have you tried TorchGeo? 🌍✨
It’s a PyTorch library built for remote sensing and geospatial data, basically torchvision for Earth observation. 🚀
With ready-to-use datasets, spatial transforms, and pre-trained models, TorchGeo makes it easy to train deep learning models on satellite imagery and geospatial layers.
🔗 Explore TorchGeo: https://github.com/torchgeo/torchgeo
#GeoTech #GIStools #RemoteSensing #SpatialAI #TorchGeo #GeospatialTech #MundusLife #EarthObservation
The new issue of our TUMcampus magazine is here! 📖Learn more about Caroline Leiss, the new head of our university library, as well as sustainable #BuildingMaterials, #spatialAI, #DigitalEducation technologies and more: https://go.tum.de/216044
📷A.Eckert
She has a passion for her job: the new head of the TUM University Library, Caroline Leiss. In this issue you can also get to know architect Jeannette Kuo, who is researching sustainable building materials. Or read an interview with Enkelejda Kasneci about how ChatGPT can lead to greater educational equity. Also in the issue: statistician Emmanuel Owiredu Odame, who came to TUM from Ghana with the TUM.Africa Talent Program to research jet lag syndrome.
#SpatialAI can enable a #drone to distinguish between static and dynamic elements. This new approach will be used by #cartographers mapping forests, in ship inspections and when building walls: http://go.tum.de/384665
📷A.Heddergott
For a good introduction to my work in SLAM, Computer Vision and Robotics, listen to my recent conversation with Gil Elbaz on the Unboxing AI podcast. Covering SLAM, neural fields, GBP, compute hardware, RL and more! Plus at the end he asked for my tips for new researchers; I hope I said useful things...
Available here (including links and transcript), or on Spotify and all the usual podcast apps:
https://t.co/bTTkyvvhfK
sayzard
Ben Lorica 罗瑞卡
Geospatial Technologies
Michael Fauscette
Technische Universität München
Andrew Davison