DepthART: Scaling Foundation Monocular Depth to Tiny Models

Benchmark Model Rank Results
monocular-depth-estimation-on-diode-relativeDepthART-L (zero-shot)#2AbsRel: 0.208delta1: 0.754
monocular-depth-estimation-on-diode-relativeDepthART-S (zero-shot)#3AbsRel: 0.214delta1: 0.745
monocular-depth-estimation-on-kitti-relativeDepthART-L (zero-shot)#6AbsRel: 0.079delta1: 0.933
monocular-depth-estimation-on-kitti-relativeDepthART-S (zero-shot)#8AbsRel: 0.082delta1: 0.930
monocular-depth-estimation-on-sun-rgbdDepthART-L (NYUD v2 fine-tuned, zero-shot)#6RMSE: 0.326
monocular-depth-estimation-on-nyuv2-relativeDepthART-L (zero-shot)#9delta1: 0.971AbsRel: 0.053
monocular-depth-estimation-on-nyuv2-relativeDepthART-S (zero-shot)#12delta1: 0.964AbsRel: 0.059
monocular-depth-estimation-on-kitti-eigen-splitDepthART-L (fine-tuned)#71Delta < 1.25: 0.956
monocular-depth-estimation-on-nyu-depth-v2DepthART-L (fine-tuned)#78Delta < 1.25: 0.946