Depth prediction (DP): takes only monocular/ stereo as input Depth completion (DC): takes depth sensor (ex: LiDAR, RADAR) as input component (converting a sparse depth map Dsparse into a dense depth map Ddense)
Metrics DP: Accu, SILog, sqErrorRel, absErrorRel, iRMSE, thresh δ
Metrics DC: Accu, iRMSE, iMAE, RMSE, MAE, thresh δ
https://papers.nips.cc/paper/2014/file/7bccfde7714a1ebadf06c5f4cea752c1-Paper.pdf
https://arxiv.org/pdf/2003.06620.pdf
http://www.cvlibs.net/datasets/kitti/eval_depth.php?benchmark=depth_prediction
| Metric | Description | Unit |
|---|---|---|
| Accuracies | ||
| SILog | Scale invariant logarithmic error  |
log(m)*100 |
| sqErrorRel | Relative squared error  |
% |
| absErrorRel | Relative absolute error |
% |
| iRMSE | Root mean squared error of the inverse depth  |
1/km |
| ● thresh δ <1.25 ● thresh δ <1.25^2 ● thresh δ <1.25^3 |
δ<threshold error provide a comprehensive comparison among the method δ<threshold error means the percent of pixels that satisfy δ<threshold, where δ  (Disp)gt: the gth disparity; (Disp)pred: the predicted disparity |
| Metric (DC) | Description | Unit |
|---|---|---|
| Accuracies | ||
| iRMSE | Root mean squared error of the inverse depth |
1/km |
| iMAE | Mean absolute error of the inverse depth  |
1/km |
| RMSE | Root mean squared error  |
mm |
| MAE | Mean absolute error  |
mm |
| δi | δi: percentage of predicted pixels where the relative error is within 1.25^i  where |.| denotes the cardinality of a set. dˆ and d are prediction and associated groundtruth. Most studies adopt i = 1,2,3 |