Table 9 Comparison with State-Of-the-Art Methods

ShapeLLM¹⁰⁰

Cross Model Framework

Yes

No

OmniVec¹⁰¹

3D object understanding

Yes

Yes

GPSFormer¹⁰²

Global Perception and Local Structure Fitting-based Transformer

No

No

TripoSR¹⁰³

Transformer architecture for fast feed-forward 3D generation

No

No

Geometry¹⁰⁴

Geometry-biased attention mechanism

Yes

Yes

UniScene¹⁰⁵

Multi-camera unified pre-training framework

Yes

No

Proposed[Vit3D]

Robust multi-scale MAE prior training architecture

Yes

Yes