flâneur — a map of the web's best reading

Shi's blog: How to Render a Single Gaussian Splat?

shi-yan.github.io · 2,836 words · saved by 1 readers

Gaussian Splatting has garnered significant attention since last year, particularly following this paper. This technique has piqued my interest due to its potential to become a standard approach for implementing 3D photography. Traditional computer graphics predominantly represent 3D scenes using triangles, and GPU hardware is optimized for rendering large quantities of triangles efficiently. The quest for reconstructing a realistic 3D scene from a set of photos taken from different angles has been a vibrant research area. However, achieving photorealism remains a challenge. Triangles fall short when it comes to representing fuzzy, transparent, or non-distinctive objects like clouds or hair. In recent years, there have been advancements in implicit 3D representations such as neural radiance fields. These methods can generate photorealistic results and handle non-distinctive objects well, albeit at the cost of intense computation and slow rendering speeds. Subsequently, researchers have

Gaussian Splatting has garnered significant attention since last year, particularly following this paper . This technique has piqued my interest due to its potential to become a standard approach for implementing 3D photography. Traditional computer graphics predominantly represent 3D scenes using triangles, and GPU hardware is optimized for rendering large quantities of triangles efficiently. The quest for reconstructing a realistic 3D scene from a set of photos taken from different angles has been a vibrant research area. However, achieving photorealism remains a challenge. Triangles fall sh

Explore this link on the map →

saved by

related reading