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

Open paper
Benchmark Model Rank Results
monocular-depth-estimation-on-scannet-relative-depth-alignedLotus-D (zero-shot)#5AbsRel: 0.055delta1: 0.965
surface-normals-estimation-on-sintelLotus-D (zero-shot)#5mAE: 32.3
monocular-depth-estimation-on-eth3d-relative-depth-alignedLotus-G (zero-shot)#6AbsRel: 0.059delta1: 0.970
monocular-depth-estimation-on-kitti-relativeLotus-D (zero-shot)#7AbsRel: 0.081delta1: 0.931
monocular-depth-estimation-on-diode-relativeLotus-D (zero-shot)#8AbsRel: 0.228delta1: 0.738
monocular-depth-estimation-on-nyuv2-relativeLotus-D (zero-shot)#8delta1: 0.972AbsRel: 0.051

Rank counts only results with a code link.