3D Gaussian Splatting vs. NeRFs. What is the difference? π€
In the world of computer vision, 3D Gaussian Splatting and NeRFs are gaining traction.
But what sets them apart? Hereβs a quick breakdown:
π 3D Space Representation:
- NeRF: Creates a continuous 3D space by sampling points throughout the scene for each training image.
- Gaussian Splatting: Relies on a sparse set of 3D points, often generated using Structure from Motion.
π¨ Points Description:
- Gaussian Splatting: Uses complex 3D forms called Gaussians with varying shapes, sizes, transparency, and color, described with spherical harmonics.
- NeRF: Assigns each point an RGBA color and a viewing direction, with appearance dependent on location and viewing angle.
βοΈ Optimization:
- NeRF: Uses a neural network to learn a continuous function for color and opacity.
- Gaussian Splatting: Directly optimizes the properties of each 3D ellipsoid without a neural network, resulting in a discrete set of ellipsoids.
In essence, NeRF offers a continuous, neural approach, while Gaussian Splatting provides a simpler, directly optimized discrete structure.
Which approach do you find more intriguing for applications in 3D graphics and AR? Share your thoughts in the comments! π¬
Information About 3D Gaussian Splatting
- Blog : https://xoft.tistory.com/74
Translate to Korean
μ»΄ν¨ν° λΉμ μ μΈκ³μμλ 3D Gaussian Splatting λ° NeRFκ° μ£Όλͺ©μ λ°κ³ μμ΅λλ€. κ·Έλ¬λ 무μμ΄ κ·Έλ€μ μ°¨λ³νν©λκΉ? λ€μμ κ°λ¨ν λΆμμ λλ€.
π 3D κ³΅κ° νν:
- NeRF: κ° νμ΅ μ΄λ―Έμ§μ λν΄ μ₯λ©΄ μ 체μ μ§μ μ μνλ§νμ¬ μ°μ 3D 곡κ°μ λ§λλλ€.
- κ°μ°μμ μ€νλν (Gaussian Splatting): μ’ μ’ λͺ¨μ μ ꡬ쑰(Structure from Motion)λ₯Ό μ¬μ©νμ¬ μμ±λλ ν¬μ 3D ν¬μΈνΈ μΈνΈλ₯Ό μ¬μ©ν©λλ€.
π¨ ν¬μΈνΈ μ€λͺ :
- κ°μ°μμ μ€νλν : ꡬν κ³ μ‘°νλ‘ μ€λͺ λλ λ€μν λͺ¨μ, ν¬κΈ°, ν¬λͺ λ λ° μμμ κ°μ§ κ°μ°μμμ΄λΌλ 볡μ‘ν 3D ννλ₯Ό μ¬μ©ν©λλ€.
- NeRF: κ° ν¬μΈνΈμ RGBA μμκ³Ό 보기 λ°©ν₯μ ν λΉνλ©°, λͺ¨μμ μμΉ λ° μμΌκ°μ λ°λΌ λ¬λΌμ§λλ€.
βοΈ μ΅μ ν:
- NeRF: μ κ²½λ§μ μ¬μ©νμ¬ μμ λ° λΆν¬λͺ λμ λν μ°μ ν¨μλ₯Ό νμ΅ν©λλ€.
- Gaussian Splatting: μ κ²½λ§ μμ΄ κ° 3D νμ체μ μμ±μ μ§μ μ΅μ ννμ¬ κ°λ³ νμ체 μΈνΈλ₯Ό μμ±ν©λλ€.
λ³Έμ§μ μΌλ‘ NeRFλ μ°μμ μΈ μ κ²½ μ κ·Ό λ°©μμ μ 곡νλ λ°λ©΄, Gaussian Splattingμ λ κ°λ¨νκ³ μ§μ μ΅μ νλ μ΄μ° ꡬ쑰λ₯Ό μ 곡ν©λλ€.
3D κ·Έλν½κ³Ό AR μ ν리μΌμ΄μ μ μ΄λ€ μ κ·Ό λ°©μμ΄ λ ν₯λ―Έλ‘λ€κ³ μκ°νμλκΉ? λκΈλ‘ μκ°μ 곡μ νμΈμ! π¬
This post is licensed under CC BY 4.0 by the author.