Objaverse

Objaverse在规模，类别数量以及类别内实例的视觉多样性方面改善了当今的3D存储库。我们通过四个不同的应用程序展示了Objaverse的巨大潜力: 训练生成的3D模型，改进LVIS基准上的尾部类别分割，训练嵌入式AI的开放词汇对象导航模型，以及为视觉模型的鲁棒性分析创建新的基准。Objaverse可以为研究开辟新的方向，并在人工智能领域实现新的应用。

Objaverse improves upon contemporary 3D repositories in terms of scale, quantity of categories, and visual diversity of instances within each category. We showcase the substantial potential of Objaverse via four distinct applications: training generative 3D models, enhancing tail-category segmentation on the LVIS benchmark, training open-vocabulary object navigation models for embedded AI, and establishing novel benchmarks for robustness analysis of vision models. Objaverse can open up new research directions and enable innovative applications within the field of artificial intelligence.