The author introduces G3DR, a novel 3D generative method in ImageNet, addressing limitations of existing methods by leveraging depth regularization and pre-trained language-vision models to improve visual realism and diversity. G3DR outperforms state-of-the-art methods by up to 22% in perceptual metrics and 90% in geometry scores.
画像から多様で高品質な3Dオブジェクトを生成する新しい方法、Generative 3D Reconstruction(G3DR)を紹介します。
G3DR bietet effiziente 3D-Asset-Generierung mit hoher Qualität und Geometrie, übertrifft den Stand der Technik um bis zu 22% in Wahrnehmungsmetriken und 90% in Geometriewerten.