Amazing example of a 3D model rendered from over 100 million 3D Gaussian distributions after scanning an entire city



Using

3D Gaussian splatting , a technique for creating 3D data from multiple images, a 3D model of an entire city has been created. The 3D model was generated on a MacBook Air with an M2 processor, and consists of approximately 106 million 3D Gaussian distributions .



The rendered 3D model of the city can be seen below.

Jastrzębia Góra 103M Splats by Andrii Shramko — splatter.app
https://splatter.app/s/lzs-xtl

The above site looks like this. You can change the camera direction by dragging the mouse, or change the viewpoint by right-clicking and moving the mouse. You can also move the viewpoint more precisely by using W (forward), S (backward), A (left), D (right), Q (up), and E (down) on the keyboard.



The camera is pulled back a little to move the viewpoint behind the building. You can see from the image that it is so intricately crafted that it is hard to believe it was rendered.



Here's a closer look at another building. Because it was rendered from an image, the texture is a bit rough, but the shape of the roof, the ducts running along the walls, and even the cars parked in the parking lot are all accurately reproduced in 3D.



The camera angle is lowered to recreate the view of the city from above. The height and shape of the buildings are accurately modeled, giving the experience a sense of realism.



If you pull the camera back as far as you can and fit the whole city into the frame, it looks like this.




The 3D model of the city was rendered from photos taken by Andriy Shramko using AirVis , an app that allows you to create 3D spaces, and a MacBook Air with M2.

Scanning and rendering with AirVis allows you to instantly create a 3D model of your surroundings.



It seems that the 3D world created by World Labs can also be explored with AirVis using Apple Vision Pro, Meta Quest 3, etc.



Related Posts:

in Video,   Software, Posted by log1i_yk