Table 5 Performance analysis of the proposed method on the KITTI dataset with different decoder structures.

From: LapUNet: a novel approach to monocular depth estimation using dynamic laplacian residual U-shape networks

Method

RMSE

RMSLE

Abs Rel

Sq Rel

δ < 1.25

δ < 1.252

δ < 1.253

Low is better

High is better

Without Lap and ASPP

2.881

0.117

0.078

0.286

0.933

0.974

0.991

Without Lap

2.573

0.108

0.071

0.248

0.947

0.982

0.997

Without ASPP

2.311

0.095

0.062

0.223

0.956

0.991

0.999

Traditional lap

2.323

0.999

0.059

0.231

0.958

0.992

0.998

Proposed method

2.247

0.085

0.055

0.200

0.964

0.994

0.999

Back to article page