The Digital Press

A depth estimation model called Depth Anything V2, developed by TikTok and the University of Hong Kong, was tested on Maxar's high-resolution satellite imagery of Bangkok to extract depth maps from urban scenes. The model performed better on smaller, cleaner image crops and shows promise for analyzing building heights and urban structures from satellite and aerial images.

Why it matters: Depth Anything V2 enables generating depth maps from satellite imagery, potentially aiding urban planning and disaster response.

The big picture: Combining satellite data with AI depth estimation could automate 3D city modeling and improve geospatial analytics.

The stakes: Model sensitivity to image quality and black areas can reduce accuracy, requiring preprocessing or image selection.

Commenters say: Users appreciate the practical application of AI on satellite imagery but note challenges with large images and the need for georeferencing.