Revealing the Dark Secrets of Masked Image Modeling

Benchmark Model Rank Results
depth-estimation-on-nyu-depth-v2SwinV2-L 1K-MIM#3RMS: 0.287
depth-estimation-on-nyu-depth-v2SwinV2-B 1K-MIM#5RMS: 0.304
monocular-depth-estimation-on-kitti-eigenSwinV2-L 1K-MIM#15absolute relative error: 0.050RMSE: 1.966Sq Rel: 0.139
monocular-depth-estimation-on-kitti-eigenSwinV2-B 1K-MIM#25absolute relative error: 0.052RMSE: 2.050Sq Rel: 0.148
monocular-depth-estimation-on-nyu-depth-v2SwinV2-L 1K-MIM#23absolute relative error: 0.083RMSE: 0.287log 10: 0.035
pose-estimation-on-coco-test-devSwinV2-L 1K-MIM#11AP: 77.2
pose-estimation-on-coco-test-devSwinV2-B 1K-MIM#14AP: 76.7
pose-estimation-on-crowdposeSwinV2-L 1K-MIM#4AP: 75.5
pose-estimation-on-crowdposeSwinV2-B 1K-MIM#5AP: 74.9
visual-object-tracking-on-got-10kSwinV2-L 1K-MIM#24Average Overlap: 72.9
visual-object-tracking-on-got-10kSwinV2-B 1K-MIM#26Average Overlap: 70.8
visual-object-tracking-on-lasotSwinV2-L 1K-MIM#26AUC: 70.7
visual-object-tracking-on-lasotSwinV2-B 1K-MIM#30AUC: 70