Lingting Zhu

MSVS-VAE: Multi-Scale Anchored VecSet for High-Fidelity 3D Reconstruction

The key idea is to progressively densify anchored VecSet latents via hierarchical point-shuffle upsampling, increasing spatial capacity for fine-grained geometry modeling and replacing global cross-attention with AVS-Conv, a geometry-aware local aggregation operator operating within local neighborhoods rather than the exhaustive latent set.