English

Evolving Evocative 2D Views of Generated 3D Objects

Computer Vision and Pattern Recognition 2021-11-10 v1

Abstract

We present a method for jointly generating 3D models of objects and 2D renders at different viewing angles, with the process guided by ImageNet and CLIP -based models. Our results indicate that it can generate anamorphic objects, with renders that both evoke the target caption and look visually appealing.

Keywords

Cite

@article{arxiv.2111.04839,
  title  = {Evolving Evocative 2D Views of Generated 3D Objects},
  author = {Eric Chu},
  journal= {arXiv preprint arXiv:2111.04839},
  year   = {2021}
}
R2 v1 2026-06-24T07:31:29.974Z