Skip to content
RobotWorld
← Back

DeWorldSG: Probabilistic 3D Gaussian Scene Graph Generation (ECCV 2026)

Loading video
Loading video

DeWorldSG generates 3D scene graphs from RGB-D using probabilistic 3D Gaussians, correcting boundary noise and depth shifts, with the V-JEPA 2 world model inferring spatial relations for robot manipulation and AR.

3D场景图Scene Graph3DGS世界模型World ModelV-JEPA感知Twitter
Category: perception
Author: @tokufxug
Date: 2026-08-14T00:00:00
Duration: 36.703s
Source: https://x.com/tokufxug/status/2088085972093477200

Related Videos

DeWorldSG: depth-aware 3D scene graph generation
1:16

DeWorldSG: depth-aware 3D scene graph generation

@rsasaki0109 · 3 views · 2026-08-13
3D场景图scene graphDeWorldSG
DJI M4E and PortalCam air-ground 3DGS map fusion
0:34

DJI M4E and PortalCam air-ground 3DGS map fusion

@DuckbillStudio · 2 views · 2026-08-07
3DGS无人机DJI Matrice 4E
NVIDIA ArtiFixer Repairs Incomplete Gaussian Splats
0:29

NVIDIA ArtiFixer Repairs Incomplete Gaussian Splats

@justinryanio · 2 views · 2026-07-20
3DGS高斯泼溅空间智能
ABot-Recon turns driving video into 3D scenes
0:16

ABot-Recon turns driving video into 3D scenes

@HuggingApps · 14 views · 2026-08-31
ABot-Recon3D重建3D reconstruction
CHC5 Open Camera: FPGA Vision Processing at 2,000 FPS
47:24

CHC5 Open Camera: FPGA Vision Processing at 2,000 FPS

@Alacritic_Super · 29 views · 2026-08-31
CHC5相机Camera
Inside SuperMap: Persistent 3D Object Identities
0:36

Inside SuperMap: Persistent 3D Object Identities

@robotsdigest · 9 views · 2026-08-16
SLAMSuperMapSuperOdometry