WebXR-based mobile point cloud rendering and semantic segmentation using depth data and voxel fusion

WebXR-based mobile point cloud rendering and semantic segmentation using depth data and voxel fusion

Researchers at National Chung Hsing University describe a mobile browser system, built on WebXR, Three.js and WebGL, that turns data from the WebXR Depth Sensing Module into point clouds and runs semantic segmentation on the device. Fusing results from several viewpoints improved accuracy in tests on a single Pixel 6 Pro running Chrome.

Source
Scientific Reports
Author
Ming-Der Yang, Chun-Wei Yeh, Hui Ping Tsai
Published
In edition
No. 1 — AI starts writing the immersive web
Sections
WebXR3D on the WebResearch & Standards

Read the full story at Scientific Reports at Scientific Reports

This page is our short summary. The full article is published by Scientific Reports; the link above takes you to their site.

Also in No. 01

WebXR3D on the Web

Introducing Three-Live: turning AI-generated Three.js into interactive 3D, AR, and WebXR across web and mobile

Google has released Three-Live, an Apache 2.0 developer preview library that compiles AI-generated Three.js code into glTF and USDZ files in a few milliseconds. On the web it runs scripts in a sandboxed Web Worker and offers an xr-live package for hand gestures, spatial interface elements and handoff into immersive WebXR sessions.

Google Open Source Blog

Original article at Google Open Source Blog

WebXR3D on the Web

Build and Stream Browser-Based XR Experiences with NVIDIA CloudXR.js

NVIDIA's CloudXR.js is a JavaScript SDK that streams GPU-rendered XR content from an OpenXR server to web browsers and headsets without an app store build. The post walks through creating a session and sending tracking data inside a WebXR render loop, and lists requirements including Meta Quest or Pico 4 Ultra headsets and Wi-Fi 6.

NVIDIA Technical Blog

Original article at NVIDIA Technical Blog