Lotus: Diffusion-based Visual Foundation Model for High-quality Dense Prediction

This model belongs to the family of official Lotus models.
Some training normals in the Hypersim dataset are not properly oriented towards the camera. This models was re-trained using aligned surface normals, referred to GeoWizard, and achieves significantly improved results.

Developed by: Jing He✱, Haodong Li✱, Wei Yin, Yixun Liang, Leheng Li, Kaiqiang Zhou, Hongbo Zhang, Bingbing Liu, Ying-Cong Chen✉

Usage

Please refer to this page.