## Method

segmentation and revision network [M2CRN]
[Anonymous Submission]

Submitted on 17 Sep. 2022 10:06 by
[Anonymous Submission]

 Running time: 0.01 s Environment: GPU @ 2.5 Ghz (Python)

 Method Description: segmentation error learning efficient road segmentation Parameters: lr=3e-4 Latex Bibtex: no

## Evaluation in Bird's Eye View

 Benchmark MaxF AP PRE REC FPR FNR UM_ROAD 96.51 % 92.57 % 96.30 % 96.72 % 1.70 % 3.28 % UMM_ROAD 97.71 % 94.81 % 97.25 % 98.18 % 3.06 % 1.82 % UU_ROAD 96.14 % 91.58 % 95.56 % 96.72 % 1.46 % 3.28 % URBAN_ROAD 96.93 % 92.93 % 96.61 % 97.26 % 1.88 % 2.74 %
This table as LaTeX

## Behavior Evaluation

 Benchmark PRE-20 F1-20 HR-20 PRE-30 F1-30 HR-30 PRE-40 F1-40 HR-40
This table as LaTeX

The following plots show precision/recall curves for the bird's eye view evaluation.

 This figure as: png eps pdf This figure as: png eps pdf This figure as: png eps pdf This figure as: png eps pdf

## Distance-dependent Behavior Evaluation

The following plots show the F1 score/Precision/Hitrate with respect to the longitudinal distance which has been used for evaluation.

## Visualization of Results

The following images illustrate the performance of the method qualitatively on a couple of test images. We first show results in the perspective image, followed by evaluation in bird's eye view. Here, red denotes false negatives, blue areas correspond to false positives and green represents true positives.

 This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png This figure as: png