1 paper · 1 filter
Junyu Li, Qiuyu Chen, Pengcheng Wang +5
Recent diffusion-based pipelines have achieved promising progress in image-to-3D synthesis. However, generating high-fidelity details remains challenging, especially when the input…