IA-project workflow part10

Image to 3D

Trellis

A novel 3D generation method is introduced, enabling versatile and high-quality 3D asset creation. At its core is a unified Structured LATent (SLAT) representation, which supports decoding into multiple output formats, including Radiance Fields, 3D Gaussians, and meshes. This is achieved by combining a sparsely-populated 3D grid with dense multiview visual features extracted from a robust vision foundation model, effectively capturing both structural (geometry) and textural (appearance) information while preserving decoding flexibility.

Structured 3D Latents for Scalable and Versatile 3D Generation (https://trellis3d.github.io/) Continuer la lecture de IA-project workflow part10

IA-project workflow part08

Better control on image generation.

In this process, it is possible to start from a basic shape or even real objects and explore results using a controlled aesthetics’ without LoRA.

Depth Calculation and visual reference

A basic 3D shape is provided along with a reference image. The shapes and style of the image will be used to the final rendering.

Continuer la lecture de IA-project workflow part08