ConOR

[ECCV 2022] Uncertainty Quantification in Depth Estimation via Constrained Ordinal Regression

Simulation Experiments

Dataset 1 (error variance)

To examine error variance estimation of ConOR, run

python main.py --cls_model conor --sample uniform --method Single

Dataset 2 (estimation variance)

To examine estimation of esimation variance with boostrap on Conor, first get ground truth by running

python main.py --cls_model conor --sample gaussian --method Truth --num_ensemble 100

Then run boostrapping methods:

python main.py --cls_model conor --sample gaussian --method MultBS --num_ensemble 100