Robbyant - Exploring the Frontiers of Embodied Intelligence | 蚂蚁灵波科技 - 探索具身智能的上限,打造物理世界的 AGI 平台

Technology-driven and application-oriented. We build foundational large models for embodied AI: spatial perception (LingBot-Depth), VLA (LingBot-VLA), world models (LingBot-World), video action (LingBot-VA). Jointly embrace the new era of embodied intelligence. 技术驱动、场景导向,自研具身智能基础大模型,共迎具身智能新时代,共创幸福生活新场景。

Robbyant 蚂蚁灵波科技

🎓 GeoTech Cohort Sharing

Proud of Ting-Jia Guo, our GeoTech Master's student, for presenting her poster at the International Summer School on LiDAR Applications & Youth Forum of DGPF at TU Darmstadt! 🏛️✈️

Her work, HERI3D, compares traditional and deep learning-based 3D reconstruction using UAV imagery to document large-scale cultural heritage sites.

#GeoTech #CohortSharing #DGPF #SummerSchool #CulturalHeritage #Photogrammetry #3DReconstruction #UAV

DeepMind and #UC #Berkeley have teamed up to give us #LoGeR, a project that claims to tackle long video 3D reconstruction. 🚀 They've thrown in buzzwords like "Hybrid Memory" and "Sliding Window Attention" to distract you from the fact that it still drifts after 19,000 frames. 😜 Basically, it's a convoluted way to say: "Look, Ma! No hands!" while tripping over the finish line. 🙃
https://loger-project.github.io #DeepMind #3DReconstruction #AIResearch #VideoTech #HackerNews #ngated
LoGeR: Long-Context Geometric Reconstruction with Hybrid Memory

Bài báo mới D4RT giới thiệu mô hình transformer duy nhất, cho phép suy đoán độ sâu, tương quan thời gian‑không gian và tham số camera từ một video duy nhất. Cơ chế truy vấn mới giảm tải tính toán, đồng thời đạt hiệu suất mới trong tái tạo 4D, vượt trội so với các phương pháp trước. #AI #ComputerVision #DeepLearning #3DReconstruction #TríTuệNhânTạo #ThịGiácMáy #TáiTạo3D

https://deepmind.google/blog/d4rt-teaching-ai-to-see-the-world-in-four-dimensions/

D4RT: Unified, Fast 4D Scene Reconstruction & Tracking

Meet D4RT, a unified AI model for 4D scene reconstruction and tracking.

Google DeepMind
Europeana’s EUreka3D is reconstructing Europe’s cultural legacy in 3D and making it interoperable across institutions.
By merging archives, models, and metadata, it’s turning scattered memory into a shared spatial archive, usable for research, education, and public storytelling.
https://pro.europeana.eu/project/eureka3d-european-union-s-rekonstructed-content-in-3d
#DigitalHeritage #3DReconstruction #CulturalMemory
EUreka3D - European Union's REKonstructed content in 3D | Europeana PRO

EUreka3D built the capacity of small cultural heritage institutions in digital transformation, particularly on issues related to 3D digitisation.

Europeana PRO
New AI model turns photos into explorable 3D worlds, with caveats

Openly available AI tool creates steerable 3D-like video, but requires serious GPU muscle.

Ars Technica

The Dispersed Chinese Art Digitization Project bridges centuries and continents through digital reconstructions of major historical sites and artifacts – from the Zhihua Temple to the Six Horses of Zhaoling and Longmen Binyang Central Cave.

Led by the Center for the Art of East Asia at UChicago and Xi’an Jiaotong University, with major museum partners in China, Japan, and the US.

https://caea.lib.uchicago.edu/dcadp/en/

#DigitalHeritage #EastAsianArt #3DReconstruction #MuseumTech #DigitalHumanities

GaVS: 3D-Grounded Video Stabilization via Temporally-Consistent Local Reconstruction
GaVS reformulates video stabilization as a 3D task: using temporally-consistent local reconstructions and Gaussian Splatting, it stabilizes full-frame video with high geometric consistency and robustness to motion and scene dynamics.
#SIGGRAPH2025 #VideoStabilization #3DReconstruction #GaussianSplatting
https://youtu.be/KNNdiv-Ea9g
https://sinoyou.github.io/gavs/
GaVS: 3D-Grounded Video Stabilization via Local Reconstruction and Rendering

YouTube
Top 7 Advances in 3D Reconstruction - Rackenzik

AI-driven 3D reconstruction enhances AR, VR, and robotics, enabling high-precision mapping and immersive digital experiences.

Rackenzik