Show HN:旧金山Sutro Tower的沉浸式高斯Splat体验
Show HN: Immersive Gaussian Splat experience of Sutro Tower, San Francisco

原始链接: https://vincentwoo.com/3d/sutro_tower/

探索我的3D模型的旧金山的Sutro塔!您可以自行速度导航,或者,如果您在电话上,请使用AR模式通过走动和使用手机来探索场景。希望您喜欢学习这种标志性的结构。要获得更深入的体验,请查看官方数字巡回演出。 该模型是使用最近的高斯剥落进步创建的。无人驾驶飞机捕获了现场,然后在现实中对齐,在GSPLAT中进行了训练,被SOGS压缩,并由PlayCanvas渲染。 Wieland Morgenstern帮助自组织高斯压缩,Donovan Hutchence有助于实施一种新的压缩格式,该格式将场景大小保持在30MB。感谢Daylen Yang驾驶第二架无人机!

这些思想集合涵盖了各种主题:由于透支和α融合而在移动设备上的高斯夹心构成了挑战,诸如基于网格的摄影测量法(Google Map,bing Maps)和Google Earth Earth VR等替代方案。 然后,讨论转向了逐步的城市发展,表明变化的速度受到控制,将新结构限制在当地建筑物密度高的百分比。这旨在避免突然的密度尖峰,例如郊区的摩天大楼和促进有机邻里进化。 辩论了城市密度的好处,包括较短的旅行距离,更有效的公用事业以及增加城市以外的绿色空间的机会。但是,承认大众运输系统的局限性,呼吁清除过时的分区以减轻压力。 最后,讨论涉及欧洲数字地面电视(DVB-T),强调了有条件的访问模块(CAM)用于付费电视频道,智能卡以解锁内容和信号加密标准。

原文

About

Welcome to my 3D model of San Francisco's Sutro Tower. Feel free to explore it at your own pace. If you're on a phone, you can also engage the AR mode by clicking the little cube, it'll let you explore the scene by walking around and waving your phone.

Sutro Tower is a wonderful building, and I hope you enjoy learning a bit about it here. If you want to learn more, check out the much more thorough official digital tour.

This scan is made possible by recent advances in Gaussian Splatting. In particular, this scene was shot on drones, aligned in RealityCapture, trained in gsplat, compressed by SOGS, and rendered by PlayCanvas.

If I've made any mistakes, or if you want to get in touch, feel free to reach out over email or Twitter.

Special Thanks

Wieland Morgenstern for developing Self Organizing Gaussian compression and assisting me in understanding it.

Donovan Hutchence of PlayCanvas for helping me implement the decoding of the new compressed format that allows us to serve this entire scene in 30MB.

Daylen Yang for flying the second drone!

Controls

Annotations

Music

Orbit Left Mouse
Touch + Drag

Pan Middle Mouse
Swipe

Look Right Mouse

Zoom Mouse Wheel
Pinch

Fly WASD
Arrow Keys

Fly faster Shift

Fly slower Ctrl

Reset Camera R

联系我们 contact @ memedata.com