## Method

RGBN RGB Fusion Net [RGBN-Fusenet]

Submitted on 18 Dec. 2019 15:53 by
Rui Fan (UC San Diego)

 Running time: 0.1 s Environment: GPU @ 2.5 Ghz (Python)

 Method Description: Fusing different types of features for semantic road surface segmentation. Parameters: lr=1e-4 Latex Bibtex:

## Evaluation in Bird's Eye View

 Benchmark MaxF AP PRE REC FPR FNR UM_ROAD 93.49 % 90.35 % 92.95 % 94.04 % 3.25 % 5.96 % UMM_ROAD 95.38 % 93.83 % 95.36 % 95.40 % 5.10 % 4.60 % UU_ROAD 91.77 % 89.46 % 91.83 % 91.72 % 2.66 % 8.28 % URBAN_ROAD 93.93 % 91.49 % 93.88 % 93.98 % 3.37 % 6.02 %
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